Images

Images are the most-tested files of all, yet generic sample sites hand you a photo with no documented properties. Here, every image records exactly what it is. Denoise, restoration, inpaint, deblur, upscale, colourization, watermark, outpaint, style-transfer, and enhance suites pair degraded inputs with clean references. Segmentation masks, visual-diff twins, EXIF/ICC ladders, multipage sprites, chart-vision, codec ladders, OCR hardcases, and soft-edge matting extend the set for modern vision QA. Clean ↔ noisy and colour ↔ greyscale pairs make before/after testing trivial.

Filter images on Browse · 2343 files · 41 subcategories

This page sorts by format. To choose by what the picture shows instead, open realism, grouped by subject.

Frequently asked questions

What makes these image fixtures different from random stock photos?

Every image ships with documented specs (dimensions, seeds, damage type, pair/group IDs). Denoise, restoration, colourization, watermark, and enhance suites include clean ground-truth twins so you can score tools with PSNR/SSIM instead of guessing.

Do you include real faces or brand watermarks?

No. Face-restore fixtures use synthetic illustrations, and watermark suites use SAMPLE logos/text only. Corrupt samples are labelled intentionally corrupt.

How do I find before/after pairs?

Open any suite member and use the Pair / Group links on the file page, or filter Browse by purpose (for example colourization-testing or watermark-removal-testing).

What vision / QA suites landed in the latest expansion?

Wave E adds segmentation scene/mask/JSON triples, A/B diff-regression twins, EXIF Orientation + ICC colour notes, multipage TIFF / sprite / favicon packs, chart-as-image mocks, codec quality ladders, hard OCR cases, and soft-alpha matting hardcases.

2343 of 2343 files

Ai Segmentation

A av f3: dense pose map

A dense pose map for the published plate a-av-f3.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av f4: dense pose map

A dense pose map for the published plate a-av-f4.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av f5: dense pose map

A dense pose map for the published plate a-av-f5.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av m3: dense pose map

A dense pose map for the published plate a-av-m3.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av m4: dense pose map

A dense pose map for the published plate a-av-m4.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av m5: dense pose map

A dense pose map for the published plate a-av-m5.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av m6: dense pose map

A dense pose map for the published plate a-av-m6.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

A av nb1: dense pose map

A dense pose map for the published plate a-av-nb1.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

png

4.8 KB

A avatar f1: dense pose map

A dense pose map for the published plate a-avatar-f1.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

png

4.8 KB

A avatar f2: dense pose map

A dense pose map for the published plate a-avatar-f2.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

png

4.5 KB

A avatar m1: dense pose map

A dense pose map for the published plate a-avatar-m1.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

png

5.2 KB

A avatar m2: dense pose map

A dense pose map for the published plate a-avatar-m2.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

Fruit Still Life on Clutter: Instances (512px)

The same four-object fruit still life on a cluttered background: the hard input for instance segmentation and object detection. The indexed instance mask gives per-object ground truth.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationObject detection· Conversion set

Fruit Still Life on White: Instances (512px)

A four-object fruit still life (apple, orange, banana, grapes) on white: the easy input for instance segmentation and object detection, paired with an indexed instance mask.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationObject detection· Conversion set

G char knight: dense pose map

A dense pose map for the published plate g-char-knight.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

G char mage: dense pose map

A dense pose map for the published plate g-char-mage.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

G char scifi: dense pose map

A dense pose map for the published plate g-char-scifi.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

Indexed Instance Mask: 4 Objects (512px)

An indexed (palette) instance mask for the fruit still life: background = 0 and each fruit painted with its own instance id. Per-object class and normalised bounding box are listed in the specs.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationObject detection· Conversion set

Invoice desk: dense pose map

A dense pose map for the published plate invoice-desk.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

Office desk: dense pose map

A dense pose map for the published plate office-desk.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

Owner cafe: dense pose map

A dense pose map for the published plate owner-cafe.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

Owner market: dense pose map

A dense pose map for the published plate owner-market.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

Owner restaurant: dense pose map

A dense pose map for the published plate owner-restaurant.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

Owner salon: dense pose map

A dense pose map for the published plate owner-salon.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

Owner shop: dense pose map

A dense pose map for the published plate owner-shop.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

Owner trades: dense pose map

A dense pose map for the published plate owner-trades.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

P accountant: dense pose map

A dense pose map for the published plate p-accountant.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P baker young: dense pose map

A dense pose map for the published plate p-baker-young.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P barista deaf: dense pose map

A dense pose map for the published plate p-barista-deaf.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P butcher mid: dense pose map

A dense pose map for the published plate p-butcher-mid.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P chef wheel: dense pose map

A dense pose map for the published plate p-chef-wheel.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P cleaner eve: dense pose map

A dense pose map for the published plate p-cleaner-eve.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P desk chair: dense pose map

A dense pose map for the published plate p-desk-chair.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

P driver eve: dense pose map

A dense pose map for the published plate p-driver-eve.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

P farmer field: dense pose map

A dense pose map for the published plate p-farmer-field.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P florist sen: dense pose map

A dense pose map for the published plate p-florist-sen.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P kitchen chef: dense pose map

A dense pose map for the published plate p-kitchen-chef.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

P market sen: dense pose map

A dense pose map for the published plate p-market-sen.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

P mechanic w: dense pose map

A dense pose map for the published plate p-mechanic-w.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P meeting mix: dense pose map

A dense pose map for the published plate p-meeting-mix.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

P nurse clinic: dense pose map

A dense pose map for the published plate p-nurse-clinic.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P salon mid: dense pose map

A dense pose map for the published plate p-salon-mid.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

P shop plus: dense pose map

A dense pose map for the published plate p-shop-plus.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

P stick cafe: dense pose map

A dense pose map for the published plate p-stick-cafe.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

P tailor mid: dense pose map

A dense pose map for the published plate p-tailor-mid.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P teacher cls: dense pose map

A dense pose map for the published plate p-teacher-cls.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

P trades w: dense pose map

A dense pose map for the published plate p-trades-w.png, 512x748. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 748 px
Use case: Pose estimation

P ware young: dense pose map

A dense pose map for the published plate p-ware-young.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

Portrait Subject: Binary Mask (512px)

The binary ground-truth mask for the portrait subject (white = foreground): the reference for scoring a portrait cut-out or background blur.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationSubject detection· Paired fixture

Portrait Subject: Input (512px)

A cat 'portrait' on a busy background: a portrait-style subject-detection and matting input, paired with its ground-truth mask.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationSubject detection+1· Paired fixture

Receipt counter: dense pose map

A dense pose map for the published plate receipt-counter.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

Salient Object: Binary Mask (512px)

The binary ground-truth segmentation mask for the salient robot (white = foreground). Compute IoU / F-score of a predicted mask against this file.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationSubject detection· Conversion set

Salient Object: Input (512px)

A robot mascot on a cluttered background: the input for a salient-object-detection or image-matting model. Ground-truth binary mask and trimap ship alongside it.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image segmentationSubject detection+1· Conversion set

Salient Object: Matting Trimap (512px)

A three-level matting trimap (black = background, grey = unknown edge band, white = foreground) for the salient robot: the standard auxiliary input for alpha-matting methods.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Image mattingImage segmentation· Conversion set

Sp climbing: dense pose map

A dense pose map for the published plate sp-climbing.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

Sp cyclists: dense pose map

A dense pose map for the published plate sp-cyclists.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

Sp football: dense pose map

A dense pose map for the published plate sp-football.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

Sp runner: dense pose map

A dense pose map for the published plate sp-runner.png, 512x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 512 × 512 px
Use case: Pose estimation

Warehouse clip: dense pose map

A dense pose map for the published plate warehouse-clip.png, 748x512. DensePose body-part segmentation, which labels regions of every person it finds rather than returning a skeleton, so background figures are labelled as well as the subject. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Segmentation · 748 × 512 px
Use case: Pose estimation

Ai Vision

A av f3: canny edge map

A canny edge map for the published plate a-av-f3.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av f3: depth map

A depth map for the published plate a-av-f3.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av f3: surface normal map

A tangent-space surface normal map for the published plate a-av-f3.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9962, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av f4: canny edge map

A canny edge map for the published plate a-av-f4.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av f4: depth map

A depth map for the published plate a-av-f4.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av f4: surface normal map

A tangent-space surface normal map for the published plate a-av-f4.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av f5: canny edge map

A canny edge map for the published plate a-av-f5.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av f5: depth map

A depth map for the published plate a-av-f5.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 87.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av f5: surface normal map

A tangent-space surface normal map for the published plate a-av-f5.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av m3: canny edge map

A canny edge map for the published plate a-av-m3.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av m3: depth map

A depth map for the published plate a-av-m3.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av m3: surface normal map

A tangent-space surface normal map for the published plate a-av-m3.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9966, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av m4: canny edge map

A canny edge map for the published plate a-av-m4.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av m4: depth map

A depth map for the published plate a-av-m4.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av m4: surface normal map

A tangent-space surface normal map for the published plate a-av-m4.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av m5: canny edge map

A canny edge map for the published plate a-av-m5.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av m5: depth map

A depth map for the published plate a-av-m5.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av m5: surface normal map

A tangent-space surface normal map for the published plate a-av-m5.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av m6: canny edge map

A canny edge map for the published plate a-av-m6.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av m6: depth map

A depth map for the published plate a-av-m6.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 94.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av m6: surface normal map

A tangent-space surface normal map for the published plate a-av-m6.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

A av nb1: canny edge map

A canny edge map for the published plate a-av-nb1.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

A av nb1: depth map

A depth map for the published plate a-av-nb1.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 87.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

A av nb1: surface normal map

A tangent-space surface normal map for the published plate a-av-nb1.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

png

49.6 KB

A avatar f1: canny edge map

A canny edge map for the published plate a-avatar-f1.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

png

179.2 KB

A avatar f1: depth map

A depth map for the published plate a-avatar-f1.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

png

465.2 KB

A avatar f1: surface normal map

A tangent-space surface normal map for the published plate a-avatar-f1.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

png

34.4 KB

A avatar f2: canny edge map

A canny edge map for the published plate a-avatar-f2.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

png

160.1 KB

A avatar f2: depth map

A depth map for the published plate a-avatar-f2.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 92.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

png

450.5 KB

A avatar f2: surface normal map

A tangent-space surface normal map for the published plate a-avatar-f2.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9968, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

png

23.9 KB

A avatar m1: canny edge map

A canny edge map for the published plate a-avatar-m1.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

png

125 KB

A avatar m1: depth map

A depth map for the published plate a-avatar-m1.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 80.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

png

433.7 KB

A avatar m1: surface normal map

A tangent-space surface normal map for the published plate a-avatar-m1.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9971, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

png

60 KB

A avatar m2: canny edge map

A canny edge map for the published plate a-avatar-m2.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

png

145.6 KB

A avatar m2: depth map

A depth map for the published plate a-avatar-m2.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

png

505.9 KB

A avatar m2: surface normal map

A tangent-space surface normal map for the published plate a-avatar-m2.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

C album abs: canny edge map

A canny edge map for the published plate c-album-abs.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

C album abs: depth map

A depth map for the published plate c-album-abs.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

C album abs: surface normal map

A tangent-space surface normal map for the published plate c-album-abs.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9969, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

C album moody: canny edge map

A canny edge map for the published plate c-album-moody.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.1% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

C album moody: depth map

A depth map for the published plate c-album-moody.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 38.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

C album moody: surface normal map

A tangent-space surface normal map for the published plate c-album-moody.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9972, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

C album retro: canny edge map

A canny edge map for the published plate c-album-retro.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 11.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

C album retro: depth map

A depth map for the published plate c-album-retro.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 96.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

C album retro: surface normal map

A tangent-space surface normal map for the published plate c-album-retro.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9945, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

C book hist: canny edge map

A canny edge map for the published plate c-book-hist.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

C book hist: depth map

A depth map for the published plate c-book-hist.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 96.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

C book hist: surface normal map

A tangent-space surface normal map for the published plate c-book-hist.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9953, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

C book nature: canny edge map

A canny edge map for the published plate c-book-nature.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

C book nature: depth map

A depth map for the published plate c-book-nature.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 81.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

C book nature: surface normal map

A tangent-space surface normal map for the published plate c-book-nature.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.997, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

C book thrill: canny edge map

A canny edge map for the published plate c-book-thrill.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

C book thrill: depth map

A depth map for the published plate c-book-thrill.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 65.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

C book thrill: surface normal map

A tangent-space surface normal map for the published plate c-book-thrill.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9954, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

D clipboard: canny edge map

A canny edge map for the published plate d-clipboard.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

D clipboard: depth map

A depth map for the published plate d-clipboard.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

D clipboard: surface normal map

A tangent-space surface normal map for the published plate d-clipboard.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.997, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

D invoice: canny edge map

A canny edge map for the published plate d-invoice.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

D invoice: depth map

A depth map for the published plate d-invoice.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

D invoice: surface normal map

A tangent-space surface normal map for the published plate d-invoice.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9949, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

D ledger: canny edge map

A canny edge map for the published plate d-ledger.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

D ledger: depth map

A depth map for the published plate d-ledger.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

D ledger: surface normal map

A tangent-space surface normal map for the published plate d-ledger.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.995, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

D menu: canny edge map

A canny edge map for the published plate d-menu.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

D menu: depth map

A depth map for the published plate d-menu.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 74.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

D menu: surface normal map

A tangent-space surface normal map for the published plate d-menu.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

D receipts: canny edge map

A canny edge map for the published plate d-receipts.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 11.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

D receipts: depth map

A depth map for the published plate d-receipts.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

D receipts: surface normal map

A tangent-space surface normal map for the published plate d-receipts.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9953, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

F bread rustic: canny edge map

A canny edge map for the published plate f-bread-rustic.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

F bread rustic: depth map

A depth map for the published plate f-bread-rustic.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 82.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

F bread rustic: surface normal map

A tangent-space surface normal map for the published plate f-bread-rustic.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9964, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

F coffee pour: canny edge map

A canny edge map for the published plate f-coffee-pour.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

F coffee pour: depth map

A depth map for the published plate f-coffee-pour.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 75.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

F coffee pour: surface normal map

A tangent-space surface normal map for the published plate f-coffee-pour.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

F plate pasta: canny edge map

A canny edge map for the published plate f-plate-pasta.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

F plate pasta: depth map

A depth map for the published plate f-plate-pasta.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

F plate pasta: surface normal map

A tangent-space surface normal map for the published plate f-plate-pasta.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9955, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

F plate salad: canny edge map

A canny edge map for the published plate f-plate-salad.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 9.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

F plate salad: depth map

A depth map for the published plate f-plate-salad.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 90.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

F plate salad: surface normal map

A tangent-space surface normal map for the published plate f-plate-salad.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G char knight: canny edge map

A canny edge map for the published plate g-char-knight.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G char knight: depth map

A depth map for the published plate g-char-knight.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G char knight: surface normal map

A tangent-space surface normal map for the published plate g-char-knight.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9964, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G char mage: canny edge map

A canny edge map for the published plate g-char-mage.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G char mage: depth map

A depth map for the published plate g-char-mage.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 78.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G char mage: surface normal map

A tangent-space surface normal map for the published plate g-char-mage.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G char scifi: canny edge map

A canny edge map for the published plate g-char-scifi.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G char scifi: depth map

A depth map for the published plate g-char-scifi.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G char scifi: surface normal map

A tangent-space surface normal map for the published plate g-char-scifi.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G dungeon: canny edge map

A canny edge map for the published plate g-dungeon.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G dungeon: depth map

A depth map for the published plate g-dungeon.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 91.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G dungeon: surface normal map

A tangent-space surface normal map for the published plate g-dungeon.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G fantasy keep: canny edge map

A canny edge map for the published plate g-fantasy-keep.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G fantasy keep: depth map

A depth map for the published plate g-fantasy-keep.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 64.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G fantasy keep: surface normal map

A tangent-space surface normal map for the published plate g-fantasy-keep.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G isometric: canny edge map

A canny edge map for the published plate g-isometric.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G isometric: depth map

A depth map for the published plate g-isometric.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 93.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G isometric: surface normal map

A tangent-space surface normal map for the published plate g-isometric.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G lowpoly isle: canny edge map

A canny edge map for the published plate g-lowpoly-isle.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G lowpoly isle: depth map

A depth map for the published plate g-lowpoly-isle.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G lowpoly isle: surface normal map

A tangent-space surface normal map for the published plate g-lowpoly-isle.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9956, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G pixel scene: canny edge map

A canny edge map for the published plate g-pixel-scene.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G pixel scene: depth map

A depth map for the published plate g-pixel-scene.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 94.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G pixel scene: surface normal map

A tangent-space surface normal map for the published plate g-pixel-scene.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9936, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G platformer: canny edge map

A canny edge map for the published plate g-platformer.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G platformer: depth map

A depth map for the published plate g-platformer.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 77.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G platformer: surface normal map

A tangent-space surface normal map for the published plate g-platformer.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G racing track: canny edge map

A canny edge map for the published plate g-racing-track.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G racing track: depth map

A depth map for the published plate g-racing-track.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 90.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G racing track: surface normal map

A tangent-space surface normal map for the published plate g-racing-track.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G scifi corr: canny edge map

A canny edge map for the published plate g-scifi-corr.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G scifi corr: depth map

A depth map for the published plate g-scifi-corr.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G scifi corr: surface normal map

A tangent-space surface normal map for the published plate g-scifi-corr.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9984, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

G voxel city: canny edge map

A canny edge map for the published plate g-voxel-city.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 10.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

G voxel city: depth map

A depth map for the published plate g-voxel-city.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

G voxel city: surface normal map

A tangent-space surface normal map for the published plate g-voxel-city.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Invoice desk: canny edge map

A canny edge map for the published plate invoice-desk.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 10.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Invoice desk: depth map

A depth map for the published plate invoice-desk.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Invoice desk: surface normal map

A tangent-space surface normal map for the published plate invoice-desk.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.995, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

L bridge: canny edge map

A canny edge map for the published plate l-bridge.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 11.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L bridge: depth map

A depth map for the published plate l-bridge.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 80.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L bridge: surface normal map

A tangent-space surface normal map for the published plate l-bridge.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L city sky: canny edge map

A canny edge map for the published plate l-city-sky.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L city sky: depth map

A depth map for the published plate l-city-sky.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 66.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L city sky: surface normal map

A tangent-space surface normal map for the published plate l-city-sky.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L coast: canny edge map

A canny edge map for the published plate l-coast.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L coast: depth map

A depth map for the published plate l-coast.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 63.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L coast: surface normal map

A tangent-space surface normal map for the published plate l-coast.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9955, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L forest path: canny edge map

A canny edge map for the published plate l-forest-path.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 12.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L forest path: depth map

A depth map for the published plate l-forest-path.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 96.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L forest path: surface normal map

A tangent-space surface normal map for the published plate l-forest-path.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L interior mod: canny edge map

A canny edge map for the published plate l-interior-mod.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L interior mod: depth map

A depth map for the published plate l-interior-mod.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 84.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L interior mod: surface normal map

A tangent-space surface normal map for the published plate l-interior-mod.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9962, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L mountain: canny edge map

A canny edge map for the published plate l-mountain.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L mountain: depth map

A depth map for the published plate l-mountain.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 69.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L mountain: surface normal map

A tangent-space surface normal map for the published plate l-mountain.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L stairwell: canny edge map

A canny edge map for the published plate l-stairwell.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L stairwell: depth map

A depth map for the published plate l-stairwell.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 96.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L stairwell: surface normal map

A tangent-space surface normal map for the published plate l-stairwell.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

L street eu: canny edge map

A canny edge map for the published plate l-street-eu.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 13.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

L street eu: depth map

A depth map for the published plate l-street-eu.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

L street eu: surface normal map

A tangent-space surface normal map for the published plate l-street-eu.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Ledger book: canny edge map

A canny edge map for the published plate ledger-book.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Ledger book: depth map

A depth map for the published plate ledger-book.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Ledger book: surface normal map

A tangent-space surface normal map for the published plate ledger-book.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9946, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Meeting table: canny edge map

A canny edge map for the published plate meeting-table.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Meeting table: depth map

A depth map for the published plate meeting-table.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 88.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Meeting table: surface normal map

A tangent-space surface normal map for the published plate meeting-table.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Menu table: canny edge map

A canny edge map for the published plate menu-table.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Menu table: depth map

A depth map for the published plate menu-table.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 92.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Menu table: surface normal map

A tangent-space surface normal map for the published plate menu-table.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

N birds feed: canny edge map

A canny edge map for the published plate n-birds-feed.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

N birds feed: depth map

A depth map for the published plate n-birds-feed.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 52.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

N birds feed: surface normal map

A tangent-space surface normal map for the published plate n-birds-feed.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.995, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

N cat window: canny edge map

A canny edge map for the published plate n-cat-window.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

N cat window: depth map

A depth map for the published plate n-cat-window.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

N cat window: surface normal map

A tangent-space surface normal map for the published plate n-cat-window.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

N dog park: canny edge map

A canny edge map for the published plate n-dog-park.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

N dog park: depth map

A depth map for the published plate n-dog-park.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 72.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

N dog park: surface normal map

A tangent-space surface normal map for the published plate n-dog-park.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

N fish tank: canny edge map

A canny edge map for the published plate n-fish-tank.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

N fish tank: depth map

A depth map for the published plate n-fish-tank.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 76.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

N fish tank: surface normal map

A tangent-space surface normal map for the published plate n-fish-tank.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9954, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

N horse field: canny edge map

A canny edge map for the published plate n-horse-field.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

N horse field: depth map

A depth map for the published plate n-horse-field.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 72.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

N horse field: surface normal map

A tangent-space surface normal map for the published plate n-horse-field.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

N sheep hill: canny edge map

A canny edge map for the published plate n-sheep-hill.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

N sheep hill: depth map

A depth map for the published plate n-sheep-hill.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 91.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

N sheep hill: surface normal map

A tangent-space surface normal map for the published plate n-sheep-hill.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9955, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Office desk: canny edge map

A canny edge map for the published plate office-desk.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Office desk: depth map

A depth map for the published plate office-desk.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Office desk: surface normal map

A tangent-space surface normal map for the published plate office-desk.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Owner cafe: canny edge map

A canny edge map for the published plate owner-cafe.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

Owner cafe: depth map

A depth map for the published plate owner-cafe.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 73.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

Owner cafe: surface normal map

A tangent-space surface normal map for the published plate owner-cafe.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9953, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

Owner market: canny edge map

A canny edge map for the published plate owner-market.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Owner market: depth map

A depth map for the published plate owner-market.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 89.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Owner market: surface normal map

A tangent-space surface normal map for the published plate owner-market.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9966, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Owner restaurant: canny edge map

A canny edge map for the published plate owner-restaurant.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

Owner restaurant: depth map

A depth map for the published plate owner-restaurant.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 93.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

Owner restaurant: surface normal map

A tangent-space surface normal map for the published plate owner-restaurant.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9976, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

Owner salon: canny edge map

A canny edge map for the published plate owner-salon.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

Owner salon: depth map

A depth map for the published plate owner-salon.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 65.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

Owner salon: surface normal map

A tangent-space surface normal map for the published plate owner-salon.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

Owner shop: canny edge map

A canny edge map for the published plate owner-shop.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

Owner shop: depth map

A depth map for the published plate owner-shop.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 90.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

Owner shop: surface normal map

A tangent-space surface normal map for the published plate owner-shop.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9968, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

Owner trades: canny edge map

A canny edge map for the published plate owner-trades.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

Owner trades: depth map

A depth map for the published plate owner-trades.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 93.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

Owner trades: surface normal map

A tangent-space surface normal map for the published plate owner-trades.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9971, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

P accountant: canny edge map

A canny edge map for the published plate p-accountant.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P accountant: depth map

A depth map for the published plate p-accountant.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 85.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P accountant: surface normal map

A tangent-space surface normal map for the published plate p-accountant.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9962, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P baker young: canny edge map

A canny edge map for the published plate p-baker-young.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P baker young: depth map

A depth map for the published plate p-baker-young.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 94.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P baker young: surface normal map

A tangent-space surface normal map for the published plate p-baker-young.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P barista deaf: canny edge map

A canny edge map for the published plate p-barista-deaf.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P barista deaf: depth map

A depth map for the published plate p-barista-deaf.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P barista deaf: surface normal map

A tangent-space surface normal map for the published plate p-barista-deaf.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P butcher mid: canny edge map

A canny edge map for the published plate p-butcher-mid.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P butcher mid: depth map

A depth map for the published plate p-butcher-mid.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 71.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P butcher mid: surface normal map

A tangent-space surface normal map for the published plate p-butcher-mid.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9956, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P chef wheel: canny edge map

A canny edge map for the published plate p-chef-wheel.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P chef wheel: depth map

A depth map for the published plate p-chef-wheel.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 89.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P chef wheel: surface normal map

A tangent-space surface normal map for the published plate p-chef-wheel.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P cleaner eve: canny edge map

A canny edge map for the published plate p-cleaner-eve.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P cleaner eve: depth map

A depth map for the published plate p-cleaner-eve.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 96.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P cleaner eve: surface normal map

A tangent-space surface normal map for the published plate p-cleaner-eve.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9969, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P desk chair: canny edge map

A canny edge map for the published plate p-desk-chair.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

P desk chair: depth map

A depth map for the published plate p-desk-chair.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 80.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

P desk chair: surface normal map

A tangent-space surface normal map for the published plate p-desk-chair.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

P driver eve: canny edge map

A canny edge map for the published plate p-driver-eve.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

P driver eve: depth map

A depth map for the published plate p-driver-eve.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

P driver eve: surface normal map

A tangent-space surface normal map for the published plate p-driver-eve.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9971, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

P farmer field: canny edge map

A canny edge map for the published plate p-farmer-field.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P farmer field: depth map

A depth map for the published plate p-farmer-field.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 71.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P farmer field: surface normal map

A tangent-space surface normal map for the published plate p-farmer-field.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P florist sen: canny edge map

A canny edge map for the published plate p-florist-sen.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P florist sen: depth map

A depth map for the published plate p-florist-sen.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 90.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P florist sen: surface normal map

A tangent-space surface normal map for the published plate p-florist-sen.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P kitchen chef: canny edge map

A canny edge map for the published plate p-kitchen-chef.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

P kitchen chef: depth map

A depth map for the published plate p-kitchen-chef.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 85.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

P kitchen chef: surface normal map

A tangent-space surface normal map for the published plate p-kitchen-chef.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

P market sen: canny edge map

A canny edge map for the published plate p-market-sen.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

P market sen: depth map

A depth map for the published plate p-market-sen.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 88.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

P market sen: surface normal map

A tangent-space surface normal map for the published plate p-market-sen.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9962, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

P mechanic w: canny edge map

A canny edge map for the published plate p-mechanic-w.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P mechanic w: depth map

A depth map for the published plate p-mechanic-w.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 87.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P mechanic w: surface normal map

A tangent-space surface normal map for the published plate p-mechanic-w.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9964, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P meeting mix: canny edge map

A canny edge map for the published plate p-meeting-mix.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

P meeting mix: depth map

A depth map for the published plate p-meeting-mix.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 91.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

P meeting mix: surface normal map

A tangent-space surface normal map for the published plate p-meeting-mix.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

P nurse clinic: canny edge map

A canny edge map for the published plate p-nurse-clinic.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P nurse clinic: depth map

A depth map for the published plate p-nurse-clinic.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 82.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P nurse clinic: surface normal map

A tangent-space surface normal map for the published plate p-nurse-clinic.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P salon mid: canny edge map

A canny edge map for the published plate p-salon-mid.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

P salon mid: depth map

A depth map for the published plate p-salon-mid.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

P salon mid: surface normal map

A tangent-space surface normal map for the published plate p-salon-mid.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9966, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

P shop plus: canny edge map

A canny edge map for the published plate p-shop-plus.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

P shop plus: depth map

A depth map for the published plate p-shop-plus.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 72.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

P shop plus: surface normal map

A tangent-space surface normal map for the published plate p-shop-plus.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

P stick cafe: canny edge map

A canny edge map for the published plate p-stick-cafe.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

P stick cafe: depth map

A depth map for the published plate p-stick-cafe.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 82.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

P stick cafe: surface normal map

A tangent-space surface normal map for the published plate p-stick-cafe.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

P tailor mid: canny edge map

A canny edge map for the published plate p-tailor-mid.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P tailor mid: depth map

A depth map for the published plate p-tailor-mid.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 87.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P tailor mid: surface normal map

A tangent-space surface normal map for the published plate p-tailor-mid.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P teacher cls: canny edge map

A canny edge map for the published plate p-teacher-cls.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

P teacher cls: depth map

A depth map for the published plate p-teacher-cls.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 94.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

P teacher cls: surface normal map

A tangent-space surface normal map for the published plate p-teacher-cls.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

P trades w: canny edge map

A canny edge map for the published plate p-trades-w.png, 832x1216. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 832 × 1216 px
Use case: Edge detection

P trades w: depth map

A depth map for the published plate p-trades-w.png, 1024x1497. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 72.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1497 px
Use case: Depth estimation

P trades w: surface normal map

A tangent-space surface normal map for the published plate p-trades-w.png, 832x1216 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 832 × 1216 px
Use case: PBR material testing

P ware young: canny edge map

A canny edge map for the published plate p-ware-young.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 9.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

P ware young: depth map

A depth map for the published plate p-ware-young.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

P ware young: surface normal map

A tangent-space surface normal map for the published plate p-ware-young.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.997, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Plate food: canny edge map

A canny edge map for the published plate plate-food.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Plate food: depth map

A depth map for the published plate plate-food.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Plate food: surface normal map

A tangent-space surface normal map for the published plate plate-food.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr backpack: canny edge map

A canny edge map for the published plate pr-backpack.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr backpack: depth map

A depth map for the published plate pr-backpack.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 95.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr backpack: surface normal map

A tangent-space surface normal map for the published plate pr-backpack.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr bottle: canny edge map

A canny edge map for the published plate pr-bottle.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr bottle: depth map

A depth map for the published plate pr-bottle.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 44.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr bottle: surface normal map

A tangent-space surface normal map for the published plate pr-bottle.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr headphone: canny edge map

A canny edge map for the published plate pr-headphone.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr headphone: depth map

A depth map for the published plate pr-headphone.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 93.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr headphone: surface normal map

A tangent-space surface normal map for the published plate pr-headphone.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9971, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr laptop: canny edge map

A canny edge map for the published plate pr-laptop.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr laptop: depth map

A depth map for the published plate pr-laptop.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 87.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr laptop: surface normal map

A tangent-space surface normal map for the published plate pr-laptop.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr plant pot: canny edge map

A canny edge map for the published plate pr-plant-pot.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr plant pot: depth map

A depth map for the published plate pr-plant-pot.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 69.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr plant pot: surface normal map

A tangent-space surface normal map for the published plate pr-plant-pot.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9956, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr shoe: canny edge map

A canny edge map for the published plate pr-shoe.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr shoe: depth map

A depth map for the published plate pr-shoe.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 52.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr shoe: surface normal map

A tangent-space surface normal map for the published plate pr-shoe.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9945, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Pr watch: canny edge map

A canny edge map for the published plate pr-watch.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Pr watch: depth map

A depth map for the published plate pr-watch.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 92.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Pr watch: surface normal map

A tangent-space surface normal map for the published plate pr-watch.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9978, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Receipt counter: canny edge map

A canny edge map for the published plate receipt-counter.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Receipt counter: depth map

A depth map for the published plate receipt-counter.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Receipt counter: surface normal map

A tangent-space surface normal map for the published plate receipt-counter.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9949, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

S bakery: canny edge map

A canny edge map for the published plate s-bakery.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 9.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

S bakery: depth map

A depth map for the published plate s-bakery.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

S bakery: surface normal map

A tangent-space surface normal map for the published plate s-bakery.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

S cafe empty: canny edge map

A canny edge map for the published plate s-cafe-empty.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S cafe empty: depth map

A depth map for the published plate s-cafe-empty.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 88.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S cafe empty: surface normal map

A tangent-space surface normal map for the published plate s-cafe-empty.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S kitchen clean: canny edge map

A canny edge map for the published plate s-kitchen-clean.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S kitchen clean: depth map

A depth map for the published plate s-kitchen-clean.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 85.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S kitchen clean: surface normal map

A tangent-space surface normal map for the published plate s-kitchen-clean.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S market stall: canny edge map

A canny edge map for the published plate s-market-stall.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 9.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S market stall: depth map

A depth map for the published plate s-market-stall.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 96.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S market stall: surface normal map

A tangent-space surface normal map for the published plate s-market-stall.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S office open: canny edge map

A canny edge map for the published plate s-office-open.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S office open: depth map

A depth map for the published plate s-office-open.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 94.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S office open: surface normal map

A tangent-space surface normal map for the published plate s-office-open.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9978, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S office: canny edge map

A canny edge map for the published plate s-office.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

S office: depth map

A depth map for the published plate s-office.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 92.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

S office: surface normal map

A tangent-space surface normal map for the published plate s-office.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9962, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

S plate: canny edge map

A canny edge map for the published plate s-plate.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S plate: depth map

A depth map for the published plate s-plate.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 92.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S plate: surface normal map

A tangent-space surface normal map for the published plate s-plate.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S shop front: canny edge map

A canny edge map for the published plate s-shop-front.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 9.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S shop front: depth map

A depth map for the published plate s-shop-front.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S shop front: surface normal map

A tangent-space surface normal map for the published plate s-shop-front.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S warehouse a: canny edge map

A canny edge map for the published plate s-warehouse-a.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 10.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

S warehouse a: depth map

A depth map for the published plate s-warehouse-a.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 93.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

S warehouse a: surface normal map

A tangent-space surface normal map for the published plate s-warehouse-a.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

S workshop: canny edge map

A canny edge map for the published plate s-workshop.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

S workshop: depth map

A depth map for the published plate s-workshop.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 91.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

S workshop: surface normal map

A tangent-space surface normal map for the published plate s-workshop.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Sp climbing: canny edge map

A canny edge map for the published plate sp-climbing.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Sp climbing: depth map

A depth map for the published plate sp-climbing.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 81.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Sp climbing: surface normal map

A tangent-space surface normal map for the published plate sp-climbing.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9974, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Sp cyclists: canny edge map

A canny edge map for the published plate sp-cyclists.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 11.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Sp cyclists: depth map

A depth map for the published plate sp-cyclists.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 93.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Sp cyclists: surface normal map

A tangent-space surface normal map for the published plate sp-cyclists.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Sp football: canny edge map

A canny edge map for the published plate sp-football.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Sp football: depth map

A depth map for the published plate sp-football.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 85.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Sp football: surface normal map

A tangent-space surface normal map for the published plate sp-football.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Sp gym: canny edge map

A canny edge map for the published plate sp-gym.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Sp gym: depth map

A depth map for the published plate sp-gym.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 88.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Sp gym: surface normal map

A tangent-space surface normal map for the published plate sp-gym.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Sp runner: canny edge map

A canny edge map for the published plate sp-runner.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

Sp runner: surface normal map

A tangent-space surface normal map for the published plate sp-runner.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Sp swimmer: depth map

A depth map for the published plate sp-swimmer.png, 1024x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 77.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Depth estimation

Sp swimmer: surface normal map

A tangent-space surface normal map for the published plate sp-swimmer.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9981, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Storefront: canny edge map

A canny edge map for the published plate storefront.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Storefront: depth map

A depth map for the published plate storefront.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 97.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Storefront: surface normal map

A tangent-space surface normal map for the published plate storefront.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9957, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

T bokeh warm: canny edge map

A canny edge map for the published plate t-bokeh-warm.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T bokeh warm: depth map

A depth map for the published plate t-bokeh-warm.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T bokeh warm: surface normal map

A tangent-space surface normal map for the published plate t-bokeh-warm.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T grad cool: canny edge map

A canny edge map for the published plate t-grad-cool.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.0% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T grad cool: depth map

A depth map for the published plate t-grad-cool.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T grad cool: surface normal map

A tangent-space surface normal map for the published plate t-grad-cool.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.997, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T grad mono: canny edge map

A canny edge map for the published plate t-grad-mono.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T grad mono: depth map

A depth map for the published plate t-grad-mono.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T grad mono: surface normal map

A tangent-space surface normal map for the published plate t-grad-mono.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T grad mono2: canny edge map

A canny edge map for the published plate t-grad-mono2.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T grad mono2: surface normal map

A tangent-space surface normal map for the published plate t-grad-mono2.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9968, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T grad warm: canny edge map

A canny edge map for the published plate t-grad-warm.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.0% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T grad warm: depth map

A depth map for the published plate t-grad-warm.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T grad warm: surface normal map

A tangent-space surface normal map for the published plate t-grad-warm.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9949, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T mesh grad: canny edge map

A canny edge map for the published plate t-mesh-grad.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T mesh grad: depth map

A depth map for the published plate t-mesh-grad.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 88.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T mesh grad: surface normal map

A tangent-space surface normal map for the published plate t-mesh-grad.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9952, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T ph dark1: canny edge map

A canny edge map for the published plate t-ph-dark1.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.1% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T ph dark1: surface normal map

A tangent-space surface normal map for the published plate t-ph-dark1.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9952, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T ph dark2: canny edge map

A canny edge map for the published plate t-ph-dark2.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T ph dark2: surface normal map

A tangent-space surface normal map for the published plate t-ph-dark2.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9965, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T ph mesh2: canny edge map

A canny edge map for the published plate t-ph-mesh2.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T ph mesh2: surface normal map

A tangent-space surface normal map for the published plate t-ph-mesh2.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9954, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T ph soft1: canny edge map

A canny edge map for the published plate t-ph-soft1.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.1% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T ph soft1: surface normal map

A tangent-space surface normal map for the published plate t-ph-soft1.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9982, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T ph soft2: canny edge map

A canny edge map for the published plate t-ph-soft2.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.0% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T ph soft2: surface normal map

A tangent-space surface normal map for the published plate t-ph-soft2.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9978, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T ph soft3: canny edge map

A canny edge map for the published plate t-ph-soft3.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 0.0% of the frame. This plate is a smooth gradient with essentially no edges, so the map is almost empty. That is the correct answer, and it ships deliberately as the degenerate case: what an edge detector returns when there is nothing to find. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T ph soft3: surface normal map

A tangent-space surface normal map for the published plate t-ph-soft3.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex brick: canny edge map

A canny edge map for the published plate t-tex-brick.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex brick: surface normal map

A tangent-space surface normal map for the published plate t-tex-brick.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9944, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex concrete: canny edge map

A canny edge map for the published plate t-tex-concrete.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex concrete: depth map

A depth map for the published plate t-tex-concrete.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 100.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T tex concrete: surface normal map

A tangent-space surface normal map for the published plate t-tex-concrete.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex denim: canny edge map

A canny edge map for the published plate t-tex-denim.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 10.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex denim: surface normal map

A tangent-space surface normal map for the published plate t-tex-denim.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9952, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex leather: canny edge map

A canny edge map for the published plate t-tex-leather.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 17.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex leather: surface normal map

A tangent-space surface normal map for the published plate t-tex-leather.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9966, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex linen: canny edge map

A canny edge map for the published plate t-tex-linen.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex linen: depth map

A depth map for the published plate t-tex-linen.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 98.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T tex linen: surface normal map

A tangent-space surface normal map for the published plate t-tex-linen.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9954, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex marble: canny edge map

A canny edge map for the published plate t-tex-marble.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 2.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex marble: surface normal map

A tangent-space surface normal map for the published plate t-tex-marble.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex paper: canny edge map

A canny edge map for the published plate t-tex-paper.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 16.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex paper: depth map

A depth map for the published plate t-tex-paper.png, 1536x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.9% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Depth estimation

T tex paper: surface normal map

A tangent-space surface normal map for the published plate t-tex-paper.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9944, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex paper2: canny edge map

A canny edge map for the published plate t-tex-paper2.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex paper2: surface normal map

A tangent-space surface normal map for the published plate t-tex-paper2.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9952, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex rust: canny edge map

A canny edge map for the published plate t-tex-rust.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 1.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex rust: surface normal map

A tangent-space surface normal map for the published plate t-tex-rust.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9967, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex sand: canny edge map

A canny edge map for the published plate t-tex-sand.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex sand: surface normal map

A tangent-space surface normal map for the published plate t-tex-sand.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9963, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex slate: canny edge map

A canny edge map for the published plate t-tex-slate.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 3.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex slate: surface normal map

A tangent-space surface normal map for the published plate t-tex-slate.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

T tex wood: canny edge map

A canny edge map for the published plate t-tex-wood.png, 1536x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 19.4% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: Edge detection

T tex wood: surface normal map

A tangent-space surface normal map for the published plate t-tex-wood.png, 1536x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9955, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1536 × 1024 px
Use case: PBR material testing

V bike city: canny edge map

A canny edge map for the published plate v-bike-city.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 12.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V bike city: surface normal map

A tangent-space surface normal map for the published plate v-bike-city.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V boat harbour: canny edge map

A canny edge map for the published plate v-boat-harbour.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V boat harbour: surface normal map

A tangent-space surface normal map for the published plate v-boat-harbour.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9956, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V bus city: canny edge map

A canny edge map for the published plate v-bus-city.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 12.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V bus city: surface normal map

A tangent-space surface normal map for the published plate v-bus-city.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9956, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V car classic: canny edge map

A canny edge map for the published plate v-car-classic.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.2% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V car classic: surface normal map

A tangent-space surface normal map for the published plate v-car-classic.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9954, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V car ev: canny edge map

A canny edge map for the published plate v-car-ev.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V car ev: surface normal map

A tangent-space surface normal map for the published plate v-car-ev.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9964, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V car night: canny edge map

A canny edge map for the published plate v-car-night.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V car night: surface normal map

A tangent-space surface normal map for the published plate v-car-night.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9952, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V car sedan: canny edge map

A canny edge map for the published plate v-car-sedan.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.8% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V car sedan: surface normal map

A tangent-space surface normal map for the published plate v-car-sedan.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9961, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V forklift: canny edge map

A canny edge map for the published plate v-forklift.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 10.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V forklift: surface normal map

A tangent-space surface normal map for the published plate v-forklift.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9973, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V moto sport: canny edge map

A canny edge map for the published plate v-moto-sport.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 6.6% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V moto sport: surface normal map

A tangent-space surface normal map for the published plate v-moto-sport.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.996, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V pickup mud: canny edge map

A canny edge map for the published plate v-pickup-mud.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.1% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V pickup mud: surface normal map

A tangent-space surface normal map for the published plate v-pickup-mud.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V plane gate: canny edge map

A canny edge map for the published plate v-plane-gate.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 5.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V plane gate: surface normal map

A tangent-space surface normal map for the published plate v-plane-gate.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9958, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V scooter: canny edge map

A canny edge map for the published plate v-scooter.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 4.7% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V scooter: surface normal map

A tangent-space surface normal map for the published plate v-scooter.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V tractor: canny edge map

A canny edge map for the published plate v-tractor.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 12.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V tractor: surface normal map

A tangent-space surface normal map for the published plate v-tractor.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9959, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V train plat: canny edge map

A canny edge map for the published plate v-train-plat.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 11.0% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V train plat: surface normal map

A tangent-space surface normal map for the published plate v-train-plat.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.997, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V truck artic: canny edge map

A canny edge map for the published plate v-truck-artic.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 8.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V truck artic: surface normal map

A tangent-space surface normal map for the published plate v-truck-artic.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9964, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

V van delivery: canny edge map

A canny edge map for the published plate v-van-delivery.png, 1024x1024. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 7.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: Edge detection

V van delivery: surface normal map

A tangent-space surface normal map for the published plate v-van-delivery.png, 1024x1024 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9955, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1024 × 1024 px
Use case: PBR material testing

Warehouse clip: canny edge map

A canny edge map for the published plate warehouse-clip.png, 1216x832. Canny edge detection at low threshold 0.1 and high 0.3, run at the plate's own resolution so the edges land on the same pixels as the photograph they came from. Lit coverage measures 9.3% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1216 × 832 px
Use case: Edge detection

Warehouse clip: depth map

A depth map for the published plate warehouse-clip.png, 1497x1024. Monocular depth from Depth Anything V2 (vitl), rendered at the plate's long edge rather than the 512-pixel default, so depth and colour can be compared per pixel without a resample in between. Lit coverage measures 99.5% of the frame. The plate and this map are the same scene at the same size, so they can be compared pixel for pixel rather than by eye.

File: PNG · Ai Vision · 1497 × 1024 px
Use case: Depth estimation

Warehouse clip: surface normal map

A tangent-space surface normal map for the published plate warehouse-clip.png, 1216x832 - the same size as the plate, so the two compare pixel for pixel with no resample in between. RGB encodes the XYZ surface direction remapped from -1..1 into 0..255. Decoded back to vectors this file measures a mean length of 0.9978, which is the number to check your own decode against: a reader that transposes the channels or inverts the remap still produces a plausible-looking image, and lights the surface the wrong way.

File: PNG · Ai Vision · 1216 × 832 px
Use case: PBR material testing

Avatar

png

820 B

Identicon Avatar: Ada Lovelace

A 256×256 identicon-style avatar generated deterministically from the seed “ada-lovelace”: a symmetric 5×5 block pattern like the default avatars on many sites. For testing avatar rendering, placeholder generation, and image pipelines.

File: PNG · Avatar
Use case: Thumbnail generationImage pipeline QA· Conversion set

png

801 B

Identicon Avatar: Alan Turing

A 256×256 identicon-style avatar generated deterministically from the seed “alan-turing”: a symmetric 5×5 block pattern like the default avatars on many sites. For testing avatar rendering, placeholder generation, and image pipelines.

File: PNG · Avatar
Use case: Thumbnail generationImage pipeline QA· Conversion set

png

822 B

Identicon Avatar: Grace Hopper

A 256×256 identicon-style avatar generated deterministically from the seed “grace-hopper”: a symmetric 5×5 block pattern like the default avatars on many sites. For testing avatar rendering, placeholder generation, and image pipelines.

File: PNG · Avatar
Use case: Thumbnail generationImage pipeline QA· Conversion set

png

810 B

Identicon Avatar: Katherine Johnson

A 256×256 identicon-style avatar generated deterministically from the seed “katherine-johnson”: a symmetric 5×5 block pattern like the default avatars on many sites. For testing avatar rendering, placeholder generation, and image pipelines.

File: PNG · Avatar
Use case: Thumbnail generationImage pipeline QA· Conversion set

Background Removal

Alpha-Channel Reveal (512px)

A left-to-right alpha gradient over a solid colour, for testing how a viewer or compositor renders partial transparency.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage pipeline QA

Apple on Cluttered Background (512px)

An apple on a busy, colour-matched background: the hard input for a background remover or matting model. Score a prediction against the paired ground-truth cut-out and alpha mask.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage segmentation· Conversion set

Apple: Ground-Truth Alpha Mask (512px)

The binary ground-truth alpha mask (white = subject, black = background) for the cluttered-background apple: compute IoU or boundary F-score of a predicted mask against this file.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage segmentation· Conversion set

Apple: Ground-Truth Cut-out (512px)

The ground-truth transparent cut-out of the apple used in the cluttered-background hard case: the ideal output of a perfect background remover.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage segmentation· Conversion set

Coffee Mug on Studio Background (512px)

A coffee mug photographed on a clean studio-white background: the easy input for a background remover, paired 1:1 with its transparent cut-out for scoring.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage pipeline QA· Paired fixture

Coffee Mug with Soft Shadow (512px)

The coffee mug with a soft drop shadow on transparency. Tests whether a background remover keeps or discards the contact shadow.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal

Coffee Mug: Transparent Cut-out (512px)

The expected transparent cut-out of the coffee mug, paired with the on-background version: the ground truth for a background-removal test.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage pipeline QA· Paired fixture

Colour-Matched Subject: Cut-out (512px)

Ground-truth cut-out for the colour-matched hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Colour-Matched Subject: Input (512px)

Subject colours deliberately match the background clutter: a hard segmentation case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Colour-Matched Subject: Mask (512px)

Binary mask for the colour-matched hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Fur-Edge Hard Case: Cut-out (512px)

Ground-truth RGBA cut-out for the fur-edge hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Fur-Edge Hard Case: Input (512px)

Synthetic fur-like fibres on a cluttered background: a classic background-removal failure case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Fur-Edge Hard Case: Mask (512px)

Binary mask for the fur-edge hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Glass Bottle Hard Case: Cut-out (512px)

Ground-truth partial-alpha glass bottle cut-out.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Glass Bottle Hard Case: Input (512px)

Semi-transparent glass bottle on clutter, partial-alpha background-removal hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Glass Bottle Hard Case: Mask (512px)

Binary mask for the glass-bottle hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Handle-Hole Hard Case: Cut-out (512px)

Ground-truth mug cut-out including the transparent handle hole.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Handle-Hole Hard Case: Input (512px)

Mug with a handle interior hole on clutter: background should show through the handle.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Handle-Hole Hard Case: Trimap (512px)

Trimap for the handle-hole mug (black/gray/white) for matting tools.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removalImage matting· Conversion set

Mesh Fence Hard Case: Cut-out (512px)

Ground-truth mesh cut-out with transparent holes.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Mesh Fence Hard Case: Input (512px)

Chain-link / mesh with interior holes. Tests whether a removers keeps background visible through the grid.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Mesh Fence Hard Case: Mask (512px)

Binary mask for the mesh-fence hard case.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal· Conversion set

Potted Plant: Fine-Edge Test (512px)

A potted plant with many thin leaves and stems on transparency: the canonical fine-edge test for how cleanly a background remover cuts around delicate foliage.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal

Semi-Transparent Drinking Glass (512px)

A half-full drinking glass rendered with partial transparency: the hard case for background removers that assume every subject pixel is opaque.

File: PNG · Background Removal · 512 × 512 px
Use case: Background removal

Business Kits

Alder Books Reconciliation: Operating chart PNG

Operating chart PNG for Alder Books Reconciliation. A 1440 by 900 RGB chart with 24 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Alder Table Bistro: Operating chart PNG

Operating chart PNG for Alder Table Bistro. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Cedar Street Tacos: Operating chart PNG

Operating chart PNG for Cedar Street Tacos. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Cobalt Operations Team: Operating chart PNG

Operating chart PNG for Cobalt Operations Team. A 1440 by 900 RGB chart with 8 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Copper Oven Bakery: Operating chart PNG

Operating chart PNG for Copper Oven Bakery. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Ember Lane Pizza: Operating chart PNG

Operating chart PNG for Ember Lane Pizza. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Fieldnote Learning Centre: Operating chart PNG

Operating chart PNG for Fieldnote Learning Centre. A 1440 by 900 RGB chart with 3 labeled source-derived observations in completions.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Fieldnote Supply Shop: Operating chart PNG

Operating chart PNG for Fieldnote Supply Shop. A 1440 by 900 RGB chart with 12 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Harbour House Stay: Operating chart PNG

Operating chart PNG for Harbour House Stay. A 1440 by 900 RGB chart with 6 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Harbour Kitchen Group: Operating chart PNG

Operating chart PNG for Harbour Kitchen Group. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Juniper Corner Cafe: Operating chart PNG

Operating chart PNG for Juniper Corner Cafe. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Juniper Fitness and Clinic Administration: Operating chart PNG

Operating chart PNG for Juniper Fitness and Clinic Administration. A 1440 by 900 RGB chart with 3 labeled source-derived observations in slots.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Juniper Workspace Cloud: Operating chart PNG

Operating chart PNG for Juniper Workspace Cloud. A 1440 by 900 RGB chart with 6 labeled source-derived observations in USD/month.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Pine Assembly Workshop: Operating chart PNG

Operating chart PNG for Pine Assembly Workshop. A 1440 by 900 RGB chart with 6 labeled source-derived observations in units.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Pine Property Services: Operating chart PNG

Operating chart PNG for Pine Property Services. A 1440 by 900 RGB chart with 12 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Tern Client Services: Operating chart PNG

Operating chart PNG for Tern Client Services. A 1440 by 900 RGB chart with 4 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Willow Community Fund: Operating chart PNG

Operating chart PNG for Willow Community Fund. A 1440 by 900 RGB chart with 3 labeled source-derived observations in USD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Willow Event Kitchen: Operating chart PNG

Operating chart PNG for Willow Event Kitchen. A 1440 by 900 RGB chart with 8 labeled source-derived observations in CAD.

File: PNG · Business Kits · 1440 × 900 px
Use case: Data import

Chart Vision

Dashboard Mock 01 (288px)

Synthetic dashboard mock (KPI cards + bar chart): SAMPLE UI for layout / chart vision models. Not a real product.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Dashboard Mock 02 (288px)

Synthetic dashboard mock (KPI cards + bar chart): SAMPLE UI for layout / chart vision models. Not a real product.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Dashboard Mock 03 (288px)

Synthetic dashboard mock (KPI cards + bar chart): SAMPLE UI for layout / chart vision models. Not a real product.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Dashboard Mock 04 (288px)

Synthetic dashboard mock (KPI cards + bar chart): SAMPLE UI for layout / chart vision models. Not a real product.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Dashboard Mock 05 (288px)

Synthetic dashboard mock (KPI cards + bar chart): SAMPLE UI for layout / chart vision models. Not a real product.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Dashboard Mock 06 (288px)

Synthetic dashboard mock (KPI cards + bar chart): SAMPLE UI for layout / chart vision models. Not a real product.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Sparkline Chart 01

Compact sparkline PNG for chart-extraction / thumbnail vision tests.

File: PNG · Chart Vision · 288 × 96 px
Use case: Computer vision· Conversion set

Sparkline Chart 02

Compact sparkline PNG for chart-extraction / thumbnail vision tests.

File: PNG · Chart Vision · 288 × 96 px
Use case: Computer vision· Conversion set

Sparkline Chart 03

Compact sparkline PNG for chart-extraction / thumbnail vision tests.

File: PNG · Chart Vision · 288 × 96 px
Use case: Computer vision· Conversion set

Table-as-Image 01 (288px)

Rendered SAMPLE data table as a PNG, for table-detection / OCR vision pipelines.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Table-as-Image 02 (288px)

Rendered SAMPLE data table as a PNG, for table-detection / OCR vision pipelines.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Table-as-Image 03 (288px)

Rendered SAMPLE data table as a PNG, for table-detection / OCR vision pipelines.

File: PNG · Chart Vision · 288 × 288 px
Use case: Computer visionOCR testing· Conversion set

Charts

Bar Chart (PNG)

A clean bar chart of monthly revenue with axis labels, gridlines, and value labels: a real-world fixture for testing chart-extraction and OCR tools, vision models, and thumbnail rendering.

File: PNG · Charts
Use case: Chart recognitionOCR testing+1· Conversion set

Line Chart (PNG)

A two-series line chart (desktop vs. mobile visitors) with markers, a legend, and gridlines, for testing chart parsing, trend extraction, and image pipelines against a known plot.

File: PNG · Charts
Use case: Chart recognitionOCR testing+1· Conversion set

Pie Chart (PNG)

A five-slice pie chart of traffic sources with percentage labels and a legend: a real-world fixture for chart-recognition, segmentation, and OCR of embedded text.

File: PNG · Charts
Use case: Chart recognitionOCR testing+1· Conversion set

Scatter Plot (PNG)

A scatter plot of 120 seeded points (order value versus items per order) showing a positive correlation, with labelled axes, for testing point-cloud extraction and chart parsing.

File: PNG · Charts
Use case: Chart recognitionThumbnail generation· Conversion set

Codec Ladder

Codec Ladder: Blocks PNG (288px)

Same blocks scene encoded as png: rung on the Wave E codec ladder for quality / convert comparisons.

File: PNG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Blocks Q20 (288px)

Same blocks scene encoded as q20: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Blocks Q50 (288px)

Same blocks scene encoded as q50: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Blocks Q80 (288px)

Same blocks scene encoded as q80: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Blocks WEBP (288px)

Same blocks scene encoded as webp: rung on the Wave E codec ladder for quality / convert comparisons.

File: WEBP · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Fruit PNG (288px)

Same fruit scene encoded as png: rung on the Wave E codec ladder for quality / convert comparisons.

File: PNG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Fruit Q20 (288px)

Same fruit scene encoded as q20: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Fruit Q50 (288px)

Same fruit scene encoded as q50: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Fruit Q80 (288px)

Same fruit scene encoded as q80: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Fruit WEBP (288px)

Same fruit scene encoded as webp: rung on the Wave E codec ladder for quality / convert comparisons.

File: WEBP · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Fruit WebP Q95 (288px)

High-quality WebP of the fruit scene, near-lossless rung of the codec ladder.

File: WEBP · Codec Ladder · 288 × 288 px
Use case: Conversion testing· Conversion set

Codec Ladder: Mug PNG (288px)

Same mug scene encoded as png: rung on the Wave E codec ladder for quality / convert comparisons.

File: PNG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Mug Q20 (288px)

Same mug scene encoded as q20: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Mug Q50 (288px)

Same mug scene encoded as q50: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Mug Q80 (288px)

Same mug scene encoded as q80: rung on the Wave E codec ladder for quality / convert comparisons.

File: JPG · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Codec Ladder: Mug WEBP (288px)

Same mug scene encoded as webp: rung on the Wave E codec ladder for quality / convert comparisons.

File: WEBP · Codec Ladder · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Color And Metadata

16-bit Grayscale PNG (deep colour)

A 16-bit (deep-colour) grayscale PNG holding a smooth 0–65535 gradient, for testing high-bit-depth support and spotting banding when a tool truncates to 8-bit.

File: PNG · Color And Metadata · 512 × 512 px
Use case: Image pipeline QAConversion testing

Baseline JPEG

The fruit still life as a baseline (sequential) JPEG, paired with a progressive JPEG of identical content for testing decode order and progressive-rendering support.

File: JPG · Color And Metadata · 512 × 512 px
Use case: Compression testingImage pipeline QA· Paired fixture

CMYK JPEG (print colour space)

The fruit still life converted to the CMYK (print) colour space and saved as a JPEG, for testing CMYK decoding and CMYK→RGB conversion. Some web viewers render CMYK JPEGs with a colour shift.

File: JPG · Color And Metadata · 512 × 512 px
Use case: Colour managementConversion testing

EXIF Metadata Present (JPEG)

A JPEG carrying rich EXIF metadata (camera make/model, software, artist, copyright, timestamp), paired with an identical metadata-stripped copy for testing EXIF reading and privacy-scrubbing tools.

File: JPG · Color And Metadata · 512 × 512 px
Use case: EXIF & orientation testingMetadata testing· Paired fixture

EXIF Metadata Stripped (JPEG)

The same photo with all EXIF metadata removed: the stripped twin of the EXIF-present JPEG, for verifying that a metadata-scrubbing tool actually removed everything.

File: JPG · Color And Metadata · 512 × 512 px
Use case: EXIF & orientation testingMetadata testing· Paired fixture

ICC sRGB-Tagged JPEG

The fruit still life saved as a JPEG with an embedded sRGB ICC colour profile, paired with an untagged copy for testing colour-management and profile handling.

File: JPG · Color And Metadata · 512 × 512 px
Use case: Colour managementImage pipeline QA· Paired fixture

PNG with Metadata Stripped

The pixel-identical PNG control with text and ICC chunks removed, for proving that a scrubber changes metadata without changing decoded RGB values.

File: PNG · Color And Metadata
Use case: Metadata testingConversion testing· Paired fixture

PNG with Text and ICC Metadata

A deterministic PNG with known text chunks and an embedded sRGB ICC profile, paired with a pixel-identical stripped control for extraction and privacy-tool tests.

File: PNG · Color And Metadata
Use case: Metadata testingConversion testing· Paired fixture

Progressive JPEG

The same photo as a progressive JPEG (loads coarse-to-fine): the twin of the baseline JPEG, for testing progressive decoding and byte-order handling.

File: JPG · Color And Metadata · 512 × 512 px
Use case: Compression testingImage pipeline QA· Paired fixture

TIFF with EXIF, XMP and IPTC Metadata

A losslessly compressed TIFF carrying known camera/software tags plus XMP and IPTC payloads, paired with an exactly decoded stripped image.

File: TIFF · Color And Metadata
Use case: Metadata testingConversion testing· Paired fixture

TIFF with Metadata Stripped

The same losslessly compressed TIFF pixels with EXIF-style, XMP and IPTC metadata absent, providing a ground-truth scrubber control.

File: TIFF · Color And Metadata
Use case: Metadata testingConversion testing· Paired fixture

Untagged (No ICC) JPEG

The same photo saved with no embedded ICC profile: the untagged twin of the sRGB-tagged JPEG, for testing how a tool assumes/handles a missing colour profile.

File: JPG · Color And Metadata · 512 × 512 px
Use case: Colour managementImage pipeline QA· Paired fixture

Colourize Set

Colourization GT: Fruit (512px)

True-colour ground truth for the fruit colourization suite. Compare a colourizer's output against this reference to score accuracy.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization GT: Mug (512px)

True-colour ground truth for the mug colourization suite. Compare a colourizer's output against this reference to score accuracy.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization GT: Plant (512px)

True-colour ground truth for the plant colourization suite. Compare a colourizer's output against this reference to score accuracy.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization Input: Fruit Greyscale (512px)

Greyscale input for colourizing the fruit scene. Pair with the colour ground truth in this group.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization Input: Fruit Sepia (512px)

Sepia input for colourizing the fruit scene. Pair with the colour ground truth in this group.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization Input: Mug Greyscale (512px)

Greyscale input for colourizing the mug scene. Pair with the colour ground truth in this group.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization Input: Mug Sepia (512px)

Sepia input for colourizing the mug scene. Pair with the colour ground truth in this group.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization Input: Plant Greyscale (512px)

Greyscale input for colourizing the plant scene. Pair with the colour ground truth in this group.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Colourization Input: Plant Sepia (512px)

Sepia input for colourizing the plant scene. Pair with the colour ground truth in this group.

File: PNG · Colourize Set · 512 × 512 px
Use case: ColourizationImage pipeline QA· Conversion set

Comfy Realism

A barista in her twenties (JPEG q90)

The same plate encoded once at JPEG quality 90. Blocking and ringing are bounded by that single pass, so a restorer can be scored against the PNG reference in this group rather than against an opinion Plate owner-cafe.

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A barista in her twenties (lossless PNG)

A synthetic photographic plate of an independent business owner, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate owner-cafe.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A Black man in his seventies (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A Black man in his seventies (lossless PNG)

A synthetic photographic plate of a person at work, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-market-sen.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A Black woman hairdresser in (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A Black woman hairdresser in (lossless PNG)

A synthetic photographic plate of a person at work, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-salon-mid.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A cappuccino with latte art in (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A cappuccino with latte art in (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate receipt-counter.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A cluttered joinery workshop (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A cluttered joinery workshop (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-workshop.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A delivery driver with a (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery driver with a (lossless PNG)

A synthetic photographic plate of a person at work, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-driver-eve.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A delivery note on a clipboard (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery note on a clipboard (lossless PNG)

A synthetic photographic plate of a photographic scene, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate d-clipboard.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A hairdresser in her forties (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A hairdresser in her forties (lossless PNG)

A synthetic photographic plate of an independent business owner, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate owner-salon.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A market stall vendor (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A market stall vendor (lossless PNG)

A synthetic photographic plate of an independent business owner, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate owner-market.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A Middle Eastern man chef (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A Middle Eastern man chef (lossless PNG)

A synthetic photographic plate of a person at work, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-kitchen-chef.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A person's hands signing a (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A person's hands signing a (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate office-desk.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A plated main course on a, plate A (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate A (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate plate-food.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate B (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate B (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-plate.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

A plus size white woman in her (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A plus size white woman in her (lossless PNG)

A synthetic photographic plate of a person at work, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-shop-plus.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate A (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate A (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate d-invoice.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate B (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate B (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate invoice-desk.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate A (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate A (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate d-menu.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate B (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate B (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate menu-table.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A scatter of crumpled paper (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A scatter of crumpled paper (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate d-receipts.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A smiling woman with short (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A smiling woman with short (lossless PNG)

A synthetic photographic plate of an independent business owner, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate owner-restaurant.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A South Asian woman (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

A South Asian woman (lossless PNG)

A synthetic photographic plate of a person at work, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-trades-w.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

A warehouse worker holding a (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A warehouse worker holding a (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate warehouse-clip.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

A wheelchair user working at (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

A wheelchair user working at (lossless PNG)

A synthetic photographic plate of a person at work, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-desk-chair.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

Abstract album cover artwork (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Abstract album cover artwork (lossless PNG)

A synthetic photographic plate of wordless square cover artwork, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate c-album-abs.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, coral (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, coral (lossless PNG)

A synthetic photographic plate of a soft low-detail background plate, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-ph-mesh2.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, violet (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, violet (lossless PNG)

A synthetic photographic plate of a flowing mesh gradient, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-mesh-grad.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Abstract warm bokeh lights out (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract warm bokeh lights out (lossless PNG)

A synthetic photographic plate of an out-of-focus light field, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-bokeh-warm.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Amateur football match on a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Amateur football match on a (lossless PNG)

A synthetic photographic plate of a sporting activity, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate sp-football.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

An East Asian man in his (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

An East Asian man in his (lossless PNG)

A synthetic photographic plate of a person at work, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-ware-young.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

An electrician in work clothes (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

An electrician in work clothes (lossless PNG)

A synthetic photographic plate of an independent business owner, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate owner-trades.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

An empty modern open plan (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

An empty modern open plan (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-office.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

An older man with a white (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

An older man with a white (lossless PNG)

A synthetic photographic plate of an independent business owner, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate owner-shop.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

An older woman with a walking (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

An older woman with a walking (lossless PNG)

A synthetic photographic plate of a person at work, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-stick-cafe.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate A (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate A (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate d-ledger.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate B (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate B (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate ledger-book.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

Arcade racing game track (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Arcade racing game track (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-racing-track.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Articulated lorry at a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Articulated lorry at a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-truck-artic.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Black woman mechanic in (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Black woman mechanic in (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-mechanic-w.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Book cover artwork, empty rain (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, empty rain (lossless PNG)

A synthetic photographic plate of wordless portrait cover artwork, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate c-book-thrill.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

Book cover artwork, misty (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, misty (lossless PNG)

A synthetic photographic plate of wordless portrait cover artwork, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate c-book-nature.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

Book cover artwork, weathered (JPEG q90)

File: JPG · Comfy Realism · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, weathered (lossless PNG)

A synthetic photographic plate of wordless portrait cover artwork, 701x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate c-book-hist.

File: PNG · Comfy Realism · 701 × 1024 px
Use case: Media handling· Conversion set

Brown horse standing in a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Brown horse standing in a (lossless PNG)

A synthetic photographic plate of an animal in a natural setting, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate n-horse-field.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Canvas backpack on a plain (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Canvas backpack on a plain (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-backpack.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

City bicycle leaning against a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

City bicycle leaning against a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-bike-city.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

City bus at a bus stop, grey (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

City bus at a bus stop, grey (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-bus-city.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

City skyline across a river at (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

City skyline across a river at (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-city-sky.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Climber on an indoor (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Climber on an indoor (lossless PNG)

A synthetic photographic plate of a sporting activity, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate sp-climbing.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up brown leather grain (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Close up brown leather grain (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-leather.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Close up indigo denim fabric (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Close up indigo denim fabric (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-denim.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Close up of coarse recycled (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coarse recycled (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-paper2.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Close up of coffee being (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up of coffee being (lossless PNG)

A synthetic photographic plate of a plated dish or food close-up, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate f-coffee-pour.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up woven linen fabric (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Close up woven linen fabric (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-linen.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Colourful stylised platformer (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Colourful stylised platformer (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-platformer.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Commuter train at a station (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Commuter train at a station (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-train-plat.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Concrete spiral stairwell seen (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Concrete spiral stairwell seen (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-stairwell.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark dungeon interior video (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark dungeon interior video (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-dungeon.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark moody blurred background, deep (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, deep (lossless PNG)

A synthetic photographic plate of a soft low-detail background plate, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-ph-dark1.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Dark moody blurred background, near (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, near (lossless PNG)

A synthetic photographic plate of a soft low-detail background plate, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-ph-dark2.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Dark saloon car on a wet city (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark saloon car on a wet city (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-car-night.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

East african woman cleaner (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

East african woman cleaner (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-cleaner-eve.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Electric scooter parked on a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Electric scooter parked on a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-scooter.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty cafe interior before (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty cafe interior before (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-cafe-empty.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty open plan office with (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty open plan office with (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-office-open.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty outdoor market stall (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty outdoor market stall (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-market-stall.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty stainless commercial (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty stainless commercial (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-kitchen-clean.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-fantasy-keep.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Filipino woman nurse at a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Filipino woman nurse at a (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-nurse-clinic.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Five colleagues of different (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

Five colleagues of different (lossless PNG)

A synthetic photographic plate of a person at work, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-meeting-mix.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

Flat concrete wall texture (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat concrete wall texture (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-concrete.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat dark slate stone surface (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat dark slate stone surface (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-slate.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat fine sand surface with (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat fine sand surface with (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-sand.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat lay of textured handmade (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat lay of textured handmade (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-paper.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat oak wood grain surface (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat oak wood grain surface (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-wood.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat red brick wall (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat red brick wall (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-brick.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat rusted steel plate (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat rusted steel plate (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-rust.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flat white marble slab with (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Flat white marble slab with (lossless PNG)

A synthetic photographic plate of a flat surface texture under even light, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-tex-marble.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Flock of sheep on a green (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Flock of sheep on a green (lossless PNG)

A synthetic photographic plate of an animal in a natural setting, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate n-sheep-hill.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Four colleagues around a (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

Four colleagues around a (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate meeting-table.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

Glass water bottle on a plain (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Glass water bottle on a plain (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-bottle.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Golden retriever sitting on (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Golden retriever sitting on (lossless PNG)

A synthetic photographic plate of an animal in a natural setting, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate n-dog-park.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Green farm tractor in a field (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Green farm tractor in a field (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-tractor.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Group of road cyclists on a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Group of road cyclists on a (lossless PNG)

A synthetic photographic plate of a sporting activity, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate sp-cyclists.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a bearded (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a bearded (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-m3.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-f3.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a middle (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a middle (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-m6.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a south (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a south (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-m4.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate A (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate A (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-m5.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate B (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate B (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-nb1.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

182.4 KB

Head and shoulders portrait of a, plate C (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

png

1.5 MB

Head and shoulders portrait of a, plate C (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-avatar-f1.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

139.4 KB

Head and shoulders portrait of a, plate D (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

png

1.3 MB

Head and shoulders portrait of a, plate D (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-avatar-f2.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

128.4 KB

Head and shoulders portrait of a, plate E (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

png

1.2 MB

Head and shoulders portrait of a, plate E (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-avatar-m1.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an east (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an east (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-f5.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an, plate A (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an, plate A (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-av-f4.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

167.5 KB

Head and shoulders portrait of an, plate B (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

png

1.4 MB

Head and shoulders portrait of an, plate B (lossless PNG)

A synthetic photographic plate of a head-and-shoulders portrait on a plain ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate a-avatar-m2.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Isometric stylised village (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Isometric stylised village (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-isometric.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Latina teacher beside a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Latina teacher beside a (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-teacher-cls.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Low poly stylised island with (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Low poly stylised island with (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-lowpoly-isle.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged indian man (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged indian man (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-accountant.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged white male butcher (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged white male butcher (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-butcher-mid.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle eastern man tailor (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle eastern man tailor (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-tailor-mid.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Modern minimalist living room (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Modern minimalist living room (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-interior-mod.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Moody album cover artwork (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Moody album cover artwork (lossless PNG)

A synthetic photographic plate of wordless square cover artwork, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate c-album-moody.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Mountain valley with a river (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Mountain valley with a river (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-mountain.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Mud splattered pickup truck on (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Mud splattered pickup truck on (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-pickup-mud.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Narrow european old town (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Narrow european old town (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-street-eu.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Older white man farmer (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Older white man farmer (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-farmer-field.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Open laptop on a plain white (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Open laptop on a plain white (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-laptop.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Over ear headphones on a plain (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Over ear headphones on a plain (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-headphone.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead colourful garden (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead colourful garden (lossless PNG)

A synthetic photographic plate of a plated dish or food close-up, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate f-plate-salad.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta (lossless PNG)

A synthetic photographic plate of a plated dish or food close-up, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate f-plate-pasta.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Passenger aircraft at an (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Passenger aircraft at an (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-plane-gate.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Path through a deciduous (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Path through a deciduous (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-forest-path.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Person lifting weights in a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Person lifting weights in a (lossless PNG)

A synthetic photographic plate of a sporting activity, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate sp-gym.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Pixel art side scrolling game (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Pixel art side scrolling game (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-pixel-scene.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Potted houseplant on a plain (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Potted houseplant on a plain (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-plant-pot.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Red vintage convertible parked (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Red vintage convertible parked (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-car-classic.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Retro album cover artwork (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Retro album cover artwork (lossless PNG)

A synthetic photographic plate of wordless square cover artwork, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate c-album-retro.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Rocky coastline with breaking (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Rocky coastline with breaking (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-coast.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Runner on a park path in the (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Runner on a park path in the (lossless PNG)

A synthetic photographic plate of a sporting activity, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate sp-runner.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Rustic sourdough loaves on a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Rustic sourdough loaves on a (lossless PNG)

A synthetic photographic plate of a plated dish or food close-up, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate f-bread-rustic.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Science fiction video game (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Science fiction video game (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-scifi-corr.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Senior south asian woman (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Senior south asian woman (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-florist-sen.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-car-sedan.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Single running shoe on a plain (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Single running shoe on a plain (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-shoe.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Small fishing boat moored in a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Small fishing boat moored in a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-boat-harbour.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Small garden birds on a wooden (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Small garden birds on a wooden (lossless PNG)

A synthetic photographic plate of an animal in a natural setting, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate n-birds-feed.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Small independent grocery (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Small independent grocery (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-shop-front.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Smooth abstract gradient (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient (lossless PNG)

A synthetic photographic plate of a smooth abstract gradient, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-grad-mono.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep (lossless PNG)

A synthetic photographic plate of a smooth abstract gradient, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-grad-cool.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Smooth abstract gradient, warm (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, warm (lossless PNG)

A synthetic photographic plate of a smooth abstract gradient, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-grad-warm.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Smooth vertical gradient from (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth vertical gradient from (lossless PNG)

A synthetic photographic plate of a smooth abstract gradient, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-grad-mono2.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Soft blurred abstract background (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background (lossless PNG)

A synthetic photographic plate of a soft low-detail background plate, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-ph-soft3.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Soft blurred abstract background, pale (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, pale (lossless PNG)

A synthetic photographic plate of a soft low-detail background plate, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-ph-soft1.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Soft blurred abstract background, warm (JPEG q90)

File: JPG · Comfy Realism · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, warm (lossless PNG)

A synthetic photographic plate of a soft low-detail background plate, 1024x683, lossless PNG. This is the reference every lossy variant in this group is derived from Plate t-ph-soft2.

File: PNG · Comfy Realism · 1024 × 683 px
Use case: Media handling· Conversion set

Sports motorcycle parked at a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Sports motorcycle parked at a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-moto-sport.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Steel truss bridge over a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Steel truss bridge over a (lossless PNG)

A synthetic photographic plate of a landscape or architectural scene, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate l-bridge.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy knight (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy knight (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-char-knight.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy mage (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy mage (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-char-mage.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised science fiction (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised science fiction (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-char-scifi.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Swimmer mid stroke in an (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Swimmer mid stroke in an (lossless PNG)

A synthetic photographic plate of a sporting activity, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate sp-swimmer.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Tabby cat sitting on a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Tabby cat sitting on a (lossless PNG)

A synthetic photographic plate of an animal in a natural setting, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate n-cat-window.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

The exterior of a small, plate A (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate A (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-bakery.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

The exterior of a small, plate B (JPEG q90)

File: JPG · Comfy Realism · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate B (lossless PNG)

A synthetic photographic plate of a photographic scene, 1024x701, lossless PNG. This is the reference every lossy variant in this group is derived from Plate storefront.

File: PNG · Comfy Realism · 1024 × 701 px
Use case: Media handling· Conversion set

Tropical fish in a planted (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Tropical fish in a planted (lossless PNG)

A synthetic photographic plate of an animal in a natural setting, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate n-fish-tank.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Voxel art city block, blocky (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Voxel art city block, blocky (lossless PNG)

A synthetic photographic plate of a stylised rendered game scene or character, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate g-voxel-city.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Warehouse aisle between tall (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Warehouse aisle between tall (lossless PNG)

A synthetic photographic plate of an interior or storefront with no people, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate s-warehouse-a.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Wheelchair user chef at a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Wheelchair user chef at a (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-chef-wheel.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

White electric hatchback at a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

White electric hatchback at a (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-car-ev.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

White panel delivery van (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

White panel delivery van (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-van-delivery.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Wristwatch on a plain grey (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Wristwatch on a plain grey (lossless PNG)

A synthetic photographic plate of a single product on a plain studio ground, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate pr-watch.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Yellow forklift in a warehouse (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Yellow forklift in a warehouse (lossless PNG)

A synthetic photographic plate of a vehicle in context, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate v-forklift.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Young east asian woman baker (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Young east asian woman baker (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-baker-young.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Young man barista signing to a (JPEG q90)

File: JPG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Young man barista signing to a (lossless PNG)

A synthetic photographic plate of a person at work, 1024x1024, lossless PNG. This is the reference every lossy variant in this group is derived from Plate p-barista-deaf.

File: PNG · Comfy Realism · 1024 × 1024 px
Use case: Media handling· Conversion set

Conditioning Maps

Conditioning Ground Truth: Apple (512px)

Colour ground-truth for the apple conditioning suite. Every map in this group is derived from this exact image, so a guided model's output can be scored against it.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningImage pipeline QA· Conversion set

Conditioning Ground Truth: Cat (512px)

Colour ground-truth for the cat conditioning suite. Every map in this group is derived from this exact image, so a guided model's output can be scored against it.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningImage pipeline QA· Conversion set

Conditioning Ground Truth: Fruit (512px)

Colour ground-truth for the fruit conditioning suite. Every map in this group is derived from this exact image, so a guided model's output can be scored against it.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningImage pipeline QA· Conversion set

Conditioning Ground Truth: Mug (512px)

Colour ground-truth for the mug conditioning suite. Every map in this group is derived from this exact image, so a guided model's output can be scored against it.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningImage pipeline QA· Conversion set

Conditioning Ground Truth: Plant (512px)

Colour ground-truth for the plant conditioning suite. Every map in this group is derived from this exact image, so a guided model's output can be scored against it.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningImage pipeline QA· Conversion set

Conditioning Ground Truth: Robot (512px)

Colour ground-truth for the robot conditioning suite. Every map in this group is derived from this exact image, so a guided model's output can be scored against it.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningImage pipeline QA· Conversion set

Conditioning Map: Apple Canny Edges (512px)

Binary Canny edge map extracted from the ground truth. Feed as a structure-control input and score the generated result against the reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningEdge detection testing+1· Conversion set

Conditioning Map: Apple Depth Map (512px)

Synthetic depth map (near = bright) for depth-conditioned generation and monocular depth-estimation regression against a known reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningDepth estimation testing+1· Conversion set

Conditioning Map: Apple Line Art (512px)

Clean line-art extraction (black lines on white) for line-art-conditioned generation and line-extraction quality tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningLine-art extraction testing+1· Conversion set

Conditioning Map: Apple Normal Map (512px)

Tangent-space surface-normal map encoded from the scene depth: a normal-conditioning input for guided generation and relighting tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningNormal map testing+1· Conversion set

Conditioning Map: Apple Scribble (512px)

Sparse scribble derived from the edges: the loose hand-drawn control input for scribble-guided image generation.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningScribble conditioning+1· Conversion set

Conditioning Map: Apple Segmentation Map (512px)

Semantic segmentation map (subject vs background): the label reference for segmentation-conditioned generation and mask-quality scoring.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningSemantic segmentation testing+1· Conversion set

Conditioning Map: Cat Canny Edges (512px)

Binary Canny edge map extracted from the ground truth. Feed as a structure-control input and score the generated result against the reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningEdge detection testing+1· Conversion set

Conditioning Map: Cat Depth Map (512px)

Synthetic depth map (near = bright) for depth-conditioned generation and monocular depth-estimation regression against a known reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningDepth estimation testing+1· Conversion set

Conditioning Map: Cat Line Art (512px)

Clean line-art extraction (black lines on white) for line-art-conditioned generation and line-extraction quality tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningLine-art extraction testing+1· Conversion set

Conditioning Map: Cat Normal Map (512px)

Tangent-space surface-normal map encoded from the scene depth: a normal-conditioning input for guided generation and relighting tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningNormal map testing+1· Conversion set

Conditioning Map: Cat Pose Skeleton (512px)

OpenPose-style keypoint skeleton for the figure: a pose-control input for guided generation and a target for pose-estimation checks.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningPose estimation testing+1· Conversion set

Conditioning Map: Cat Scribble (512px)

Sparse scribble derived from the edges: the loose hand-drawn control input for scribble-guided image generation.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningScribble conditioning+1· Conversion set

Conditioning Map: Cat Segmentation Map (512px)

Semantic segmentation map (subject vs background): the label reference for segmentation-conditioned generation and mask-quality scoring.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningSemantic segmentation testing+1· Conversion set

Conditioning Map: Fruit Canny Edges (512px)

Binary Canny edge map extracted from the ground truth. Feed as a structure-control input and score the generated result against the reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningEdge detection testing+1· Conversion set

Conditioning Map: Fruit Depth Map (512px)

Synthetic depth map (near = bright) for depth-conditioned generation and monocular depth-estimation regression against a known reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningDepth estimation testing+1· Conversion set

Conditioning Map: Fruit Line Art (512px)

Clean line-art extraction (black lines on white) for line-art-conditioned generation and line-extraction quality tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningLine-art extraction testing+1· Conversion set

Conditioning Map: Fruit Normal Map (512px)

Tangent-space surface-normal map encoded from the scene depth: a normal-conditioning input for guided generation and relighting tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningNormal map testing+1· Conversion set

Conditioning Map: Fruit Scribble (512px)

Sparse scribble derived from the edges: the loose hand-drawn control input for scribble-guided image generation.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningScribble conditioning+1· Conversion set

Conditioning Map: Fruit Segmentation Map (512px)

Semantic segmentation map (subject vs background): the label reference for segmentation-conditioned generation and mask-quality scoring.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningSemantic segmentation testing+1· Conversion set

Conditioning Map: Mug Canny Edges (512px)

Binary Canny edge map extracted from the ground truth. Feed as a structure-control input and score the generated result against the reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningEdge detection testing+1· Conversion set

Conditioning Map: Mug Depth Map (512px)

Synthetic depth map (near = bright) for depth-conditioned generation and monocular depth-estimation regression against a known reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningDepth estimation testing+1· Conversion set

Conditioning Map: Mug Line Art (512px)

Clean line-art extraction (black lines on white) for line-art-conditioned generation and line-extraction quality tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningLine-art extraction testing+1· Conversion set

Conditioning Map: Mug Normal Map (512px)

Tangent-space surface-normal map encoded from the scene depth: a normal-conditioning input for guided generation and relighting tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningNormal map testing+1· Conversion set

Conditioning Map: Mug Scribble (512px)

Sparse scribble derived from the edges: the loose hand-drawn control input for scribble-guided image generation.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningScribble conditioning+1· Conversion set

Conditioning Map: Mug Segmentation Map (512px)

Semantic segmentation map (subject vs background): the label reference for segmentation-conditioned generation and mask-quality scoring.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningSemantic segmentation testing+1· Conversion set

Conditioning Map: Plant Canny Edges (512px)

Binary Canny edge map extracted from the ground truth. Feed as a structure-control input and score the generated result against the reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningEdge detection testing+1· Conversion set

Conditioning Map: Plant Depth Map (512px)

Synthetic depth map (near = bright) for depth-conditioned generation and monocular depth-estimation regression against a known reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningDepth estimation testing+1· Conversion set

Conditioning Map: Plant Line Art (512px)

Clean line-art extraction (black lines on white) for line-art-conditioned generation and line-extraction quality tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningLine-art extraction testing+1· Conversion set

Conditioning Map: Plant Normal Map (512px)

Tangent-space surface-normal map encoded from the scene depth: a normal-conditioning input for guided generation and relighting tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningNormal map testing+1· Conversion set

Conditioning Map: Plant Scribble (512px)

Sparse scribble derived from the edges: the loose hand-drawn control input for scribble-guided image generation.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningScribble conditioning+1· Conversion set

Conditioning Map: Plant Segmentation Map (512px)

Semantic segmentation map (subject vs background): the label reference for segmentation-conditioned generation and mask-quality scoring.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningSemantic segmentation testing+1· Conversion set

Conditioning Map: Robot Canny Edges (512px)

Binary Canny edge map extracted from the ground truth. Feed as a structure-control input and score the generated result against the reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningEdge detection testing+1· Conversion set

Conditioning Map: Robot Depth Map (512px)

Synthetic depth map (near = bright) for depth-conditioned generation and monocular depth-estimation regression against a known reference.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningDepth estimation testing+1· Conversion set

Conditioning Map: Robot Line Art (512px)

Clean line-art extraction (black lines on white) for line-art-conditioned generation and line-extraction quality tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningLine-art extraction testing+1· Conversion set

Conditioning Map: Robot Normal Map (512px)

Tangent-space surface-normal map encoded from the scene depth: a normal-conditioning input for guided generation and relighting tests.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningNormal map testing+1· Conversion set

Conditioning Map: Robot Pose Skeleton (512px)

OpenPose-style keypoint skeleton for the figure: a pose-control input for guided generation and a target for pose-estimation checks.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningPose estimation testing+1· Conversion set

Conditioning Map: Robot Scribble (512px)

Sparse scribble derived from the edges: the loose hand-drawn control input for scribble-guided image generation.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningScribble conditioning+1· Conversion set

Conditioning Map: Robot Segmentation Map (512px)

Semantic segmentation map (subject vs background): the label reference for segmentation-conditioned generation and mask-quality scoring.

File: PNG · Conditioning Maps · 512 × 512 px
Use case: ControlNet conditioningSemantic segmentation testing+1· Conversion set

Deblur Set

Deblur Base A: Clean (512px)

Clean reference for deblur base A. Compare deblurred outputs of the blurred variants in this group against this file.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Base B: Clean (512px)

Clean reference for deblur base B. Compare deblurred outputs of the blurred variants in this group against this file.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Gauss 3: Base A (512px)

Base A with gauss-3 blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Gauss 3: Base B (512px)

Base B with gauss-3 blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Gauss 6: Base A (512px)

Base A with gauss-6 blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Gauss 6: Base B (512px)

Base B with gauss-6 blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Motion Diag: Base A (512px)

Base A with motion-diag blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Motion Diag: Base B (512px)

Base B with motion-diag blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Motion H: Base A (512px)

Base A with motion-h blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Motion H: Base B (512px)

Base B with motion-h blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Motion V: Base A (512px)

Base A with motion-v blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Motion V: Base B (512px)

Base B with motion-v blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Shake: Base A (512px)

Base A with shake blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Deblur Shake: Base B (512px)

Base B with shake blur. Score a deblur tool against the clean reference in this group.

File: PNG · Deblur Set · 512 × 512 px
Use case: Deblur testingImage pipeline QA· Conversion set

Denoise Set

Denoise Base A: Clean Reference (512px)

The clean reference image for denoise base A: a fruit still life on studio white. Diff the noisy variants in this set against this file to measure filter accuracy.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Denoise Base B: Clean Reference (512px)

The clean reference image for denoise base B: a coffee mug on studio white. Diff the noisy variants in this set against this file to measure filter accuracy.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Gaussian Noise σ=10: Base A (512px)

Base A with additive Gaussian noise at σ=10 (seed 4210). Compare against the clean reference in this set to score a denoise filter.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Gaussian Noise σ=10: Base B (512px)

Base B with additive Gaussian noise at σ=10 (seed 7710). Compare against the clean reference in this set to score a denoise filter.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Gaussian Noise σ=25: Base A (512px)

Base A with additive Gaussian noise at σ=25 (seed 4225). Compare against the clean reference in this set to score a denoise filter.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Gaussian Noise σ=25: Base B (512px)

Base B with additive Gaussian noise at σ=25 (seed 7725). Compare against the clean reference in this set to score a denoise filter.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Gaussian Noise σ=50: Base A (512px)

Base A with additive Gaussian noise at σ=50 (seed 4250). Compare against the clean reference in this set to score a denoise filter.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Gaussian Noise σ=50: Base B (512px)

Base B with additive Gaussian noise at σ=50 (seed 7750). Compare against the clean reference in this set to score a denoise filter.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Salt & Pepper 10%: Base A (512px)

Base A with 10% salt-and-pepper noise (seed 42010). A direct fixture for testing median and morphological denoisers against the clean reference.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Salt & Pepper 10%: Base B (512px)

Base B with 10% salt-and-pepper noise (seed 77010). A direct fixture for testing median and morphological denoisers against the clean reference.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Salt & Pepper 2%: Base A (512px)

Base A with 2% salt-and-pepper noise (seed 42002). A direct fixture for testing median and morphological denoisers against the clean reference.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Salt & Pepper 2%: Base B (512px)

Base B with 2% salt-and-pepper noise (seed 77002). A direct fixture for testing median and morphological denoisers against the clean reference.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Salt & Pepper 5%: Base A (512px)

Base A with 5% salt-and-pepper noise (seed 42005). A direct fixture for testing median and morphological denoisers against the clean reference.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Salt & Pepper 5%: Base B (512px)

Base B with 5% salt-and-pepper noise (seed 77005). A direct fixture for testing median and morphological denoisers against the clean reference.

File: PNG · Denoise Set · 512 × 512 px
Use case: Denoise testingImage pipeline QA· Conversion set

Diff Regression

Diff Regression A: 1Px Crop (288px)

Baseline twin A for visual-diff regression (pixel change suite: 1px-crop). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: 1Px Shift (288px)

Baseline twin A for visual-diff regression (pixel change suite: 1px-shift). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: Button Add (288px)

Baseline twin A for visual-diff regression (button change suite: button-add). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: Button Label (288px)

Baseline twin A for visual-diff regression (button change suite: button-label). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: Colour Tint (288px)

Baseline twin A for visual-diff regression (colour change suite: colour-tint). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: Corner Badge (288px)

Baseline twin A for visual-diff regression (button change suite: corner-badge). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: Dot Noise (288px)

Baseline twin A for visual-diff regression (pixel change suite: dot-noise). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression A: Hue Shift (288px)

Baseline twin A for visual-diff regression (colour change suite: hue-shift). Compare against twin B.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: 1Px Crop (288px)

Changed twin B for visual-diff regression (pixel: 1px-crop). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: 1Px Shift (288px)

Changed twin B for visual-diff regression (pixel: 1px-shift). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: Button Add (288px)

Changed twin B for visual-diff regression (button: button-add). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: Button Label (288px)

Changed twin B for visual-diff regression (button: button-label). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: Colour Tint (288px)

Changed twin B for visual-diff regression (colour: colour-tint). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: Corner Badge (288px)

Changed twin B for visual-diff regression (button: corner-badge). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: Dot Noise (288px)

Changed twin B for visual-diff regression (pixel: dot-noise). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Diff Regression B: Hue Shift (288px)

Changed twin B for visual-diff regression (colour: hue-shift). Expect a non-zero pixel / perceptual diff vs A.

File: PNG · Diff Regression · 288 × 288 px
Use case: Visual diff / regression· Paired fixture

Edge Cases

Animated GIF (8 frames)

An 8-frame looping animated GIF of a moving dot. Tests animation handling, first-frame extraction, and GIF parsing.

File: GIF · Edge Cases · 200 × 200 px
Use case: Image pipeline QA

EXIF Orientation 1 (normal)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 1 (normal). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 2 (mirror horizontal)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 2 (mirror horizontal). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 3 (rotate 180°)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 3 (rotate 180°). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 4 (mirror vertical)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 4 (mirror vertical). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 5 (mirror + rotate 90° CCW)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 5 (mirror + rotate 90° CCW). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 6 (rotate 90° CW)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 6 (rotate 90° CW). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 7 (mirror + rotate 90° CW)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 7 (mirror + rotate 90° CW). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

EXIF Orientation 8 (rotate 90° CCW)

A JPEG whose pixels are stored identically but whose EXIF Orientation tag is set to 8 (rotate 90° CCW). Tests whether your tool honours orientation metadata.

File: JPG · Edge Cases · 400 × 300 px
Use case: EXIF & orientation testing

Extreme Aspect Strip (1×5000)

A 1-pixel-wide, 5000-pixel-tall strip: an extreme aspect ratio for stress-testing layout, scaling, and thumbnail generators.

File: PNG · Edge Cases · 1 × 5000 px
Use case: Image pipeline QA

Intentionally Corrupt: Truncated JPEG

An intentionally corrupt JPEG, truncated to half its bytes, for testing how a decoder handles incomplete image data. This is not a valid image by design.

File: JPG · Edge Cases · 400 × 300 px
Use case: Error handling

JPEG Quality 10 (512px)

The same 512px source saved at JPEG quality 10: part of a q10/50/90 ladder for comparing compression artefacts at a glance.

File: JPG · Edge Cases · 512 × 512 px
Use case: Compression testingConversion testing· Conversion set

JPEG Quality 50 (512px)

The same 512px source saved at JPEG quality 50: part of a q10/50/90 ladder for comparing compression artefacts at a glance.

File: JPG · Edge Cases · 512 × 512 px
Use case: Compression testingConversion testing· Conversion set

JPEG Quality 90 (512px)

The same 512px source saved at JPEG quality 90: part of a q10/50/90 ladder for comparing compression artefacts at a glance.

File: JPG · Edge Cases · 512 × 512 px
Use case: Compression testingConversion testing· Conversion set

Enhance Set

Enhancement Clean Reference: Fruit (512px)

Clean fruit still life for a second enhancement ladder subject.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Clean Reference: Mug (512px)

Correctly exposed mug reference for scoring auto-enhance, exposure correction, and dehaze tools.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Cool Cast Mug (512px)

Mug photo with deliberate cool cast for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Haze Mug (512px)

Mug photo with deliberate haze for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Hazy Fruit (512px)

Hazy fruit scene for dehaze / clarity enhancement tests.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: High Contrast Mug (512px)

Mug photo with deliberate high contrast for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Low Contrast Mug (512px)

Mug photo with deliberate low contrast for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Overexposed Mug (512px)

Mug photo with deliberate overexposed for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Underexposed Fruit (512px)

Severely underexposed fruit scene for enhancement / brighten tools.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Underexposed Mug (512px)

Mug photo with deliberate underexposed for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

Enhancement Input: Warm Cast Mug (512px)

Mug photo with deliberate warm cast for testing auto-enhance and colour/exposure correction against the clean reference.

File: PNG · Enhance Set · 512 × 512 px
Use case: Image enhancementImage pipeline QA· Conversion set

EXIF Icc

EXIF Orientation: Flip H (tag 2)

JPEG with EXIF Orientation=2 (flip-h). Pixels are upright: viewers/tools should apply the tagged transform. SAMPLE fixture.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Flip V (tag 4)

JPEG with EXIF Orientation=4 (flip-v). Pixels are upright: viewers/tools should apply the tagged transform. SAMPLE fixture.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Normal (tag 1)

JPEG with EXIF Orientation=1 (normal). Pixels are upright: viewers/tools should apply the tagged transform. SAMPLE fixture.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Rotate 180 (tag 3)

JPEG with EXIF Orientation=3 (rotate-180). Pixels are upright: viewers/tools should apply the tagged transform. SAMPLE fixture.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Rotate 90 Ccw (tag 8)

JPEG with EXIF Orientation=8 (rotate-90-ccw). Pixels are upright: viewers/tools should apply the tagged transform. SAMPLE fixture.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Rotate 90 Cw (tag 6)

JPEG with EXIF Orientation=6 (rotate-90-cw). Pixels are upright: viewers/tools should apply the tagged transform. SAMPLE fixture.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Transpose (tag 5)

JPEG with less-common EXIF Orientation=5 (transpose) for exhaustive orientation-handler coverage.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

EXIF Orientation: Transverse (tag 7)

JPEG with less-common EXIF Orientation=7 (transverse) for exhaustive orientation-handler coverage.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profileConversion testing· Conversion set

ICC sRGB-Tagged JPEG (Wave E)

Mug scene JPEG with an embedded sRGB ICC profile. Pair with the untagged twin for colour-management tests.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profile· Paired fixture

sRGB Colour-Space Note PNG

PNG whose colour intent is documented as sRGB in specs (no mandatory ICC blob), for tools that read metadata/specs.

File: PNG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profile· Conversion set

Untagged (No ICC) JPEG (Wave E)

Identical mug JPEG with no ICC profile. Specs note that sRGB is commonly assumed when untagged.

File: JPG · EXIF Icc · 288 × 288 px
Use case: ICC / colour profile· Paired fixture

Face Restore

Face Restore Clean: Synthetic Blur Base (512px)

Clean synthetic face illustration (not a real person) used as ground truth for face-restore / enhancement tools.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Clean: Synthetic Combo Base (512px)

Clean synthetic face illustration (not a real person) used as ground truth for face-restore / enhancement tools.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Clean: Synthetic Downscale Base (512px)

Clean synthetic face illustration (not a real person) used as ground truth for face-restore / enhancement tools.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Clean: Synthetic Jpeg-Crush Base (512px)

Clean synthetic face illustration (not a real person) used as ground truth for face-restore / enhancement tools.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Damaged: Blur (512px)

Deliberately degraded synthetic face (blur) for testing face restore / enhancement against the clean reference.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Damaged: Combo (512px)

Deliberately degraded synthetic face (combo) for testing face restore / enhancement against the clean reference.

File: JPG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Damaged: Downscale (512px)

Deliberately degraded synthetic face (down) for testing face restore / enhancement against the clean reference.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Damaged: Jpeg-Crush (512px)

Deliberately degraded synthetic face (jpeg) for testing face restore / enhancement against the clean reference.

File: JPG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Mask: Blur (512px)

Damage-region mask (L) for the face-restore case: white = degraded region.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Mask: Combo (512px)

Damage-region mask (L) for the face-restore case: white = degraded region.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Mask: Downscale (512px)

Damage-region mask (L) for the face-restore case: white = degraded region.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Face Restore Mask: Jpeg-Crush (512px)

Damage-region mask (L) for the face-restore case: white = degraded region.

File: PNG · Face Restore · 512 × 512 px
Use case: Face restoreImage pipeline QA· Conversion set

Format Set

Format Set: Fruit Still Life (AVIF)

A 1024×1024 fruit still life exported as AVIF: the modern-codec member of the conversion set.

File: AVIF · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (BMP)

A 1024×1024 fruit still life exported as BMP: one member of a conversion set that carries identical content across formats, so you can convert one and diff against the others.

File: BMP · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (GIF)

A 1024×1024 fruit still life exported as GIF: one member of a conversion set that carries identical content across formats, so you can convert one and diff against the others.

File: GIF · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (HEIC)

A 1024×1024 fruit still life exported as HEIC: the HEIF/HEVC member of the conversion set, the format modern phones use for photos.

File: HEIC · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (HEIF)

A 1024×1024 fruit still life exported as HEIF: the HEIF/HEVC member of the conversion set, the format modern phones use for photos.

File: HEIF · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (JPG)

A 1024×1024 fruit still life exported as JPG: one member of a conversion set that carries identical content across formats, so you can convert one and diff against the others.

File: JPG · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (PNG)

A 1024×1024 fruit still life exported as PNG: one member of a conversion set that carries identical content across formats, so you can convert one and diff against the others.

File: PNG · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (TIFF)

A 1024×1024 fruit still life exported as TIFF: one member of a conversion set that carries identical content across formats, so you can convert one and diff against the others.

File: TIFF · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Format Set: Fruit Still Life (WEBP)

A 1024×1024 fruit still life exported as WEBP: one member of a conversion set that carries identical content across formats, so you can convert one and diff against the others.

File: WEBP · Format Set · 1024 × 1024 px
Use case: Conversion testing· Conversion set

Greyscale Set

16-Step Grey Wedge

A 16-step greyscale wedge from pure black to pure white, for testing tone reproduction, banding, and monitor calibration.

File: PNG · Greyscale Set · 512 × 128 px
Use case: Greyscale testingImage pipeline QA

50% Grey Card

A flat 50% grey card (rgb 128,128,128), a reference for white balance and exposure testing.

File: PNG · Greyscale Set · 256 × 256 px
Use case: Greyscale testingImage pipeline QA

Black & White Checkerboard

A pure black-and-white checkerboard, useful for testing scaling, aliasing, and edge handling.

File: PNG · Greyscale Set · 512 × 512 px
Use case: Image pipeline QA

Black-to-White Tonal Ramp

A continuous linear greyscale ramp from black to white, useful for spotting banding and gamma issues.

File: PNG · Greyscale Set · 512 × 64 px
Use case: Greyscale testing

Fruit Still Life: Colour (512px)

A colour fruit still life (apple, orange, banana, grapes), paired with a true-greyscale conversion: a real-world subject for testing desaturation and how distinct hues collapse to similar greys.

File: PNG · Greyscale Set · 512 × 512 px
Use case: Greyscale testingImage pipeline QA· Paired fixture

Fruit Still Life: Greyscale (512px)

True 8-bit greyscale conversion of the fruit still life, the paired reference for testing colour-to-grey conversion.

File: PNG · Greyscale Set · 512 × 512 px
Use case: Greyscale testingImage pipeline QA· Paired fixture

Full-Saturation Hue Bars: Colour (512px)

Vertical bars sweeping the full hue circle at maximum saturation, paired with a greyscale version that shows how colours of different hues collapse to similar greys.

File: PNG · Greyscale Set · 512 × 256 px
Use case: Greyscale testingImage pipeline QA· Paired fixture

Full-Saturation Hue Bars: Greyscale (512px)

True 8-bit greyscale conversion of the full-saturation hue bars, the paired reference for testing colour-to-grey conversion.

File: PNG · Greyscale Set · 512 × 256 px
Use case: Greyscale testingImage pipeline QA· Paired fixture

Gradient Composition: Colour (512px)

A smooth three-channel gradient composition, paired with its greyscale conversion to test luminance handling.

File: PNG · Greyscale Set · 512 × 512 px
Use case: Greyscale testingImage pipeline QA· Paired fixture

Gradient Composition: Greyscale (512px)

True 8-bit greyscale conversion of the gradient composition, the paired reference for testing colour-to-grey conversion.

File: PNG · Greyscale Set · 512 × 512 px
Use case: Greyscale testingImage pipeline QA· Paired fixture

Hdr Linear

8-bit Greyscale Twin (PNG)

8-bit PNG twin of the linear 16-bit greyscale ramp. Compare banding after truncation.

File: PNG · Hdr Linear · 256 × 256 px
Use case: Linear 16-bit / HDR-ishConversion testing· Conversion set

Linear 16-bit Checker TIFF

64×64 linear 16-bit checkerboard TIFF for bit-depth and downsample tests.

File: TIFF · Hdr Linear · 64 × 64 px
Use case: Linear 16-bit / HDR-ishImage pipeline QA· Conversion set

Linear 16-bit Grey TIFF (deflate)

64×64 linear 16-bit greyscale TIFF with deflate: Wave H depth for high bit-depth loaders.

File: TIFF · Hdr Linear · 64 × 64 px
Use case: Linear 16-bit / HDR-ishImage pipeline QA+1· Conversion set

Linear 16-bit Grey TIFF (LZW)

LZW twin of the Wave H linear 16-bit greyscale TIFF.

File: TIFF · Hdr Linear · 64 × 64 px
Use case: Linear 16-bit / HDR-ishConversion testing· Conversion set

Linear 16-bit Greyscale TIFF

A linear 16-bit greyscale TIFF ramp (0–65535). Not a PQ HDR master, for testing high bit-depth TIFF loaders and 16→8 truncation.

File: TIFF · Hdr Linear · 256 × 256 px
Use case: Linear 16-bit / HDR-ishImage pipeline QA+1· Conversion set

Linear 16-bit Highlight Clip TIFF

A 16-bit TIFF biased into the upper code values (near clip), for highlight-rolloff and tone-map tests on linear data.

File: TIFF · Hdr Linear · 256 × 256 px
Use case: Linear 16-bit / HDR-ishImage pipeline QA· Conversion set

Linear 16-bit Hot-Edge TIFF

Linear 16-bit TIFF with a dark edge and near-clip field, tone-map edge case.

File: TIFF · Hdr Linear · 64 × 64 px
Use case: Linear 16-bit / HDR-ishImage pipeline QA· Conversion set

Inpaint Cuts

A av f3: clean plate (ground truth)

A clean 512x512 crop of head and shoulders portrait of a black woman with locs, plain terracotta studio background, taken from the published plate a-av-f3.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av f3: object composited in

The same view of head and shoulders portrait of a black woman with locs, plain terracotta studio background, with a foreign object composited over 6.15% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av f3: object removed

Head and shoulders portrait of a black woman with locs, plain terracotta studio background, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 38.409/255 and from the ground-truth plate by 5.719/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 3.157/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av f3: removal mask

The exact footprint of the object sitting over head and shoulders portrait of a black woman with locs, plain terracotta studio background, as an 8-bit mask covering 6.15% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av nb1: clean plate (ground truth)

A clean 512x512 crop of head and shoulders portrait of a young androgynous person with short pink hair, plain lilac background, taken from the published plate a-av-nb1.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av nb1: object composited in

The same view of head and shoulders portrait of a young androgynous person with short pink hair, plain lilac background, with a foreign object composited over 5.08% of the frame as one long thin region, where growing the mask by 12 pixels more than doubles the area it covers. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av nb1: object removed

Head and shoulders portrait of a young androgynous person with short pink hair, plain lilac background, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 51.809/255 and from the ground-truth plate by 7.091/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 2.995/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

A av nb1: removal mask

The exact footprint of the object sitting over head and shoulders portrait of a young androgynous person with short pink hair, plain lilac background, as an 8-bit mask covering 5.08% of the frame as one long thin region, where growing the mask by 12 pixels more than doubles the area it covers. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

png

359.3 KB

A avatar m1: clean plate (ground truth)

A clean 512x512 crop of head and shoulders portrait of a smiling black man against a plain grey studio background, taken from the published plate a-avatar-m1.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

png

364.7 KB

A avatar m1: object composited in

The same view of head and shoulders portrait of a smiling black man against a plain grey studio background, with a foreign object composited over 5.29% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

png

349.8 KB

A avatar m1: object removed

Head and shoulders portrait of a smiling black man against a plain grey studio background, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 54.279/255 and from the ground-truth plate by 10.905/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.019/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

png

1.1 KB

A avatar m1: removal mask

The exact footprint of the object sitting over head and shoulders portrait of a smiling black man against a plain grey studio background, as an 8-bit mask covering 5.29% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

C album moody: clean plate (ground truth)

A clean 512x512 crop of moody album cover artwork, lone figure silhouette against fog, muted palette, taken from the published plate c-album-moody.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

C album moody: object composited in

The same view of moody album cover artwork, lone figure silhouette against fog, muted palette, with a foreign object composited over 10.96% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

C album moody: object removed

Moody album cover artwork, lone figure silhouette against fog, muted palette, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 46.628/255 and from the ground-truth plate by 1.685/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 2.574/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

C album moody: removal mask

The exact footprint of the object sitting over moody album cover artwork, lone figure silhouette against fog, muted palette, as an 8-bit mask covering 10.96% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

D ledger: clean plate (ground truth)

A clean 512x512 crop of an open paper accounts ledger with ruled columns, fountain pen resting on the page, window light, taken from the published plate d-ledger.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

D ledger: object composited in

The same view of an open paper accounts ledger with ruled columns, fountain pen resting on the page, window light, with a foreign object composited over 5.60% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

D ledger: object removed

An open paper accounts ledger with ruled columns, fountain pen resting on the page, window light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 106.971/255 and from the ground-truth plate by 21.952/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 3.353/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

D ledger: removal mask

The exact footprint of the object sitting over an open paper accounts ledger with ruled columns, fountain pen resting on the page, window light, as an 8-bit mask covering 5.60% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G char knight: clean plate (ground truth)

A clean 512x512 crop of stylised fantasy knight character render, full body, neutral grey backdrop, taken from the published plate g-char-knight.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G char knight: object composited in

The same view of stylised fantasy knight character render, full body, neutral grey backdrop, with a foreign object composited over 10.66% of the frame as one long thin region, where growing the mask by 12 pixels more than doubles the area it covers. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G char knight: object removed

Stylised fantasy knight character render, full body, neutral grey backdrop, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 52.182/255 and from the ground-truth plate by 20.634/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 6.795/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G char knight: removal mask

The exact footprint of the object sitting over stylised fantasy knight character render, full body, neutral grey backdrop, as an 8-bit mask covering 10.66% of the frame as one long thin region, where growing the mask by 12 pixels more than doubles the area it covers. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G dungeon: clean plate (ground truth)

A clean 512x512 crop of dark dungeon interior video game environment, torchlight, stone arches, taken from the published plate g-dungeon.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G dungeon: object composited in

The same view of dark dungeon interior video game environment, torchlight, stone arches, with a foreign object composited over 15.76% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G dungeon: object removed

Dark dungeon interior video game environment, torchlight, stone arches, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 62.511/255 and from the ground-truth plate by 24.609/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 7.771/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G dungeon: removal mask

The exact footprint of the object sitting over dark dungeon interior video game environment, torchlight, stone arches, as an 8-bit mask covering 15.76% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G lowpoly isle: clean plate (ground truth)

A clean 512x512 crop of low poly stylised island with trees and water, taken from the published plate g-lowpoly-isle.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G lowpoly isle: object composited in

The same view of low poly stylised island with trees and water, with a foreign object composited over 6.50% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G lowpoly isle: object removed

Low poly stylised island with trees and water, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 55.589/255 and from the ground-truth plate by 75.537/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 5.011/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G lowpoly isle: removal mask

The exact footprint of the object sitting over low poly stylised island with trees and water, as an 8-bit mask covering 6.50% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G racing track: clean plate (ground truth)

A clean 512x512 crop of arcade racing game track through a canyon, motion blur, taken from the published plate g-racing-track.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G racing track: object composited in

The same view of arcade racing game track through a canyon, motion blur, with a foreign object composited over 15.07% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G racing track: object removed

Arcade racing game track through a canyon, motion blur, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 32.313/255 and from the ground-truth plate by 12.923/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.956/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G racing track: removal mask

The exact footprint of the object sitting over arcade racing game track through a canyon, motion blur, as an 8-bit mask covering 15.07% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G scifi corr: clean plate (ground truth)

A clean 512x512 crop of science fiction video game corridor, emissive panels, volumetric light, taken from the published plate g-scifi-corr.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G scifi corr: object composited in

The same view of science fiction video game corridor, emissive panels, volumetric light, with a foreign object composited over 10.89% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G scifi corr: object removed

Science fiction video game corridor, emissive panels, volumetric light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 53.065/255 and from the ground-truth plate by 38.959/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 3.258/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

G scifi corr: removal mask

The exact footprint of the object sitting over science fiction video game corridor, emissive panels, volumetric light, as an 8-bit mask covering 10.89% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Clean Reference (512px)

Clean fruit still-life reference for the inpaint cut suite. Each cut variant and mask in this group points back to this file.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Centre Circle (512px)

Missing-region (centre-circle) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Centre Rect (512px)

Missing-region (centre-rect) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Corner (512px)

Missing-region (corner) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Edge Crop (512px)

Missing-region (edge-crop) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Irregular Tear (512px)

Missing-region (irregular-tear) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Transparent Hole (512px)

Missing-region (transparent-hole) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Cut: Vertical Strip (512px)

Missing-region (vertical-strip) input for inpainting / content-aware fill tests. Use with the matching mask and clean reference.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: InpaintingPhoto restoration· Conversion set

Inpaint Mask: Centre Circle (512px)

White = missing region for the centre-circle cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Mask: Centre Rect (512px)

White = missing region for the centre-rect cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Mask: Corner (512px)

White = missing region for the corner cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Mask: Edge Crop (512px)

White = missing region for the edge-crop cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Mask: Irregular Tear (512px)

White = missing region for the irregular-tear cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Mask: Transparent Hole (512px)

White = missing region for the transparent-hole cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Inpaint Mask: Vertical Strip (512px)

White = missing region for the vertical-strip cut. Feed this mask with the cut input to an inpainting tool.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L bridge: clean plate (ground truth)

A clean 512x512 crop of steel truss bridge over a river, clear daylight, taken from the published plate l-bridge.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L bridge: object composited in

The same view of steel truss bridge over a river, clear daylight, with a foreign object composited over 7.24% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L bridge: object removed

Steel truss bridge over a river, clear daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 65.894/255 and from the ground-truth plate by 41.698/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 8.333/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L bridge: removal mask

The exact footprint of the object sitting over steel truss bridge over a river, clear daylight, as an 8-bit mask covering 7.24% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L stairwell: clean plate (ground truth)

A clean 512x512 crop of concrete spiral stairwell seen from below, strong geometry, daylight, taken from the published plate l-stairwell.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L stairwell: object composited in

The same view of concrete spiral stairwell seen from below, strong geometry, daylight, with a foreign object composited over 11.73% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L stairwell: object removed

Concrete spiral stairwell seen from below, strong geometry, daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 65.853/255 and from the ground-truth plate by 31.839/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 5.063/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L stairwell: removal mask

The exact footprint of the object sitting over concrete spiral stairwell seen from below, strong geometry, daylight, as an 8-bit mask covering 11.73% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L street eu: clean plate (ground truth)

A clean 512x512 crop of narrow european old town street, cobbles, receding , daylight, taken from the published plate l-street-eu.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L street eu: object composited in

The same view of narrow european old town street, cobbles, receding , daylight, with a foreign object composited over 9.47% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L street eu: object removed

Narrow european old town street, cobbles, receding , daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 62.106/255 and from the ground-truth plate by 38.069/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 11.016/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

L street eu: removal mask

The exact footprint of the object sitting over narrow european old town street, cobbles, receding , daylight, as an 8-bit mask covering 9.47% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Office desk: clean plate (ground truth)

A clean 512x512 crop of a person's hands signing a document at a desk, pen held over spread paperwork, a wristwatch visible at the cuff, taken from the published plate office-desk.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Office desk: object composited in

The same view of a person's hands signing a document at a desk, pen held over spread paperwork, a wristwatch visible at the cuff, with a foreign object composited over 4.54% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Office desk: object removed

A person's hands signing a document at a desk, pen held over spread paperwork, a wristwatch visible at the cuff, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 158.044/255 and from the ground-truth plate by 7.06/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 3.463/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Office desk: removal mask

The exact footprint of the object sitting over a person's hands signing a document at a desk, pen held over spread paperwork, a wristwatch visible at the cuff, as an 8-bit mask covering 4.54% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner restaurant: clean plate (ground truth)

A clean 512x512 crop of a smiling woman with short grey hair in a grey apron, arms folded, standing at a bistro counter below chalkboard menus with bread baskets behind her, taken from the published plate owner-restaurant.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner restaurant: object composited in

The same view of a smiling woman with short grey hair in a grey apron, arms folded, standing at a bistro counter below chalkboard menus with bread baskets behind her, with a foreign object composited over 10.86% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner restaurant: object removed

A smiling woman with short grey hair in a grey apron, arms folded, standing at a bistro counter below chalkboard menus with bread baskets behind her, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 118.767/255 and from the ground-truth plate by 12.881/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 14.477/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner restaurant: removal mask

The exact footprint of the object sitting over a smiling woman with short grey hair in a grey apron, arms folded, standing at a bistro counter below chalkboard menus with bread baskets behind her, as an 8-bit mask covering 10.86% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner trades: clean plate (ground truth)

A clean 512x512 crop of an electrician in work clothes checking a consumer unit in a domestic hallway, tool bag open on the floor, natural light, taken from the published plate owner-trades.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner trades: object composited in

The same view of an electrician in work clothes checking a consumer unit in a domestic hallway, tool bag open on the floor, natural light, with a foreign object composited over 4.99% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner trades: object removed

An electrician in work clothes checking a consumer unit in a domestic hallway, tool bag open on the floor, natural light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 72.598/255 and from the ground-truth plate by 18.235/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.619/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Owner trades: removal mask

The exact footprint of the object sitting over an electrician in work clothes checking a consumer unit in a domestic hallway, tool bag open on the floor, natural light, as an 8-bit mask covering 4.99% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P butcher mid: clean plate (ground truth)

A clean 512x512 crop of middle aged white male butcher in striped apron behind a counter, hanging scales, tiled wall, taken from the published plate p-butcher-mid.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P butcher mid: object composited in

The same view of middle aged white male butcher in striped apron behind a counter, hanging scales, tiled wall, with a foreign object composited over 4.61% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P butcher mid: object removed

Middle aged white male butcher in striped apron behind a counter, hanging scales, tiled wall, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 43.785/255 and from the ground-truth plate by 23.557/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 6.11/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P butcher mid: removal mask

The exact footprint of the object sitting over middle aged white male butcher in striped apron behind a counter, hanging scales, tiled wall, as an 8-bit mask covering 4.61% of the frame as a region touching the frame, so the inpainter has no surrounding context on one side. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P desk chair: clean plate (ground truth)

A clean 512x512 crop of a wheelchair user working at an accessible office desk with two monitors and printed papers, open plan office, soft daylight, taken from the published plate p-desk-chair.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P desk chair: object composited in

The same view of a wheelchair user working at an accessible office desk with two monitors and printed papers, open plan office, soft daylight, with a foreign object composited over 9.28% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P desk chair: object removed

A wheelchair user working at an accessible office desk with two monitors and printed papers, open plan office, soft daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 65.106/255 and from the ground-truth plate by 59.559/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 5.553/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P desk chair: removal mask

The exact footprint of the object sitting over a wheelchair user working at an accessible office desk with two monitors and printed papers, open plan office, soft daylight, as an 8-bit mask covering 9.28% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P florist sen: clean plate (ground truth)

A clean 512x512 crop of senior south asian woman florist wrapping stems in paper, bucket of blooms, shopfront light, taken from the published plate p-florist-sen.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P florist sen: object composited in

The same view of senior south asian woman florist wrapping stems in paper, bucket of blooms, shopfront light, with a foreign object composited over 5.02% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P florist sen: object removed

Senior south asian woman florist wrapping stems in paper, bucket of blooms, shopfront light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 48.638/255 and from the ground-truth plate by 39.582/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 5.552/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P florist sen: removal mask

The exact footprint of the object sitting over senior south asian woman florist wrapping stems in paper, bucket of blooms, shopfront light, as an 8-bit mask covering 5.02% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P mechanic w: clean plate (ground truth)

A clean 512x512 crop of black woman mechanic in overalls under a raised car, work lamp, garage interior, taken from the published plate p-mechanic-w.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P mechanic w: object composited in

The same view of black woman mechanic in overalls under a raised car, work lamp, garage interior, with a foreign object composited over 10.24% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P mechanic w: object removed

Black woman mechanic in overalls under a raised car, work lamp, garage interior, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 49.413/255 and from the ground-truth plate by 57.367/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 7.292/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P mechanic w: removal mask

The exact footprint of the object sitting over black woman mechanic in overalls under a raised car, work lamp, garage interior, as an 8-bit mask covering 10.24% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P meeting mix: clean plate (ground truth)

A clean 512x512 crop of five colleagues of different ages and ethnicities around a meeting table with printed reports and a laptop, office daylight, taken from the published plate p-meeting-mix.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P meeting mix: object composited in

The same view of five colleagues of different ages and ethnicities around a meeting table with printed reports and a laptop, office daylight, with a foreign object composited over 16.00% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P meeting mix: object removed

Five colleagues of different ages and ethnicities around a meeting table with printed reports and a laptop, office daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 82.621/255 and from the ground-truth plate by 56.593/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 6.934/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P meeting mix: removal mask

The exact footprint of the object sitting over five colleagues of different ages and ethnicities around a meeting table with printed reports and a laptop, office daylight, as an 8-bit mask covering 16.00% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P shop plus: clean plate (ground truth)

A clean 512x512 crop of a plus size white woman in her fifties behind the counter of a small hardware shop, shelves of tools behind her, shopfront daylight, taken from the published plate p-shop-plus.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P shop plus: object composited in

The same view of a plus size white woman in her fifties behind the counter of a small hardware shop, shelves of tools behind her, shopfront daylight, with a foreign object composited over 7.87% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P shop plus: object removed

A plus size white woman in her fifties behind the counter of a small hardware shop, shelves of tools behind her, shopfront daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 47.083/255 and from the ground-truth plate by 59.823/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.116/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P shop plus: removal mask

The exact footprint of the object sitting over a plus size white woman in her fifties behind the counter of a small hardware shop, shelves of tools behind her, shopfront daylight, as an 8-bit mask covering 7.87% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P teacher cls: clean plate (ground truth)

A clean 512x512 crop of latina teacher beside a whiteboard talking to seated adult students, classroom, window light, taken from the published plate p-teacher-cls.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P teacher cls: object composited in

The same view of latina teacher beside a whiteboard talking to seated adult students, classroom, window light, with a foreign object composited over 16.78% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P teacher cls: object removed

Latina teacher beside a whiteboard talking to seated adult students, classroom, window light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 43.944/255 and from the ground-truth plate by 42.892/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.889/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

P teacher cls: removal mask

The exact footprint of the object sitting over latina teacher beside a whiteboard talking to seated adult students, classroom, window light, as an 8-bit mask covering 16.78% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr backpack: clean plate (ground truth)

A clean 512x512 crop of canvas backpack on a plain grey studio background, three quarter, soft light, taken from the published plate pr-backpack.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr backpack: object composited in

The same view of canvas backpack on a plain grey studio background, three quarter, soft light, with a foreign object composited over 15.42% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr backpack: object removed

Canvas backpack on a plain grey studio background, three quarter, soft light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 53.202/255 and from the ground-truth plate by 43.265/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 9.491/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr backpack: removal mask

The exact footprint of the object sitting over canvas backpack on a plain grey studio background, three quarter, soft light, as an 8-bit mask covering 15.42% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr laptop: clean plate (ground truth)

A clean 512x512 crop of open laptop on a plain white studio background, three quarter, soft light, blank screen, taken from the published plate pr-laptop.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr laptop: object composited in

The same view of open laptop on a plain white studio background, three quarter, soft light, blank screen, with a foreign object composited over 8.33% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr laptop: object removed

Open laptop on a plain white studio background, three quarter, soft light, blank screen, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 53.18/255 and from the ground-truth plate by 4.38/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 10.44/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr laptop: removal mask

The exact footprint of the object sitting over open laptop on a plain white studio background, three quarter, soft light, blank screen, as an 8-bit mask covering 8.33% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr watch: clean plate (ground truth)

A clean 512x512 crop of wristwatch on a plain grey studio background, soft light, taken from the published plate pr-watch.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr watch: object composited in

The same view of wristwatch on a plain grey studio background, soft light, with a foreign object composited over 9.31% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr watch: object removed

Wristwatch on a plain grey studio background, soft light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 35.125/255 and from the ground-truth plate by 7.094/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 8.446/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Pr watch: removal mask

The exact footprint of the object sitting over wristwatch on a plain grey studio background, soft light, as an 8-bit mask covering 9.31% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S kitchen clean: clean plate (ground truth)

A clean 512x512 crop of empty stainless commercial kitchen, clean surfaces, overhead lighting, wide angle, taken from the published plate s-kitchen-clean.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S kitchen clean: object composited in

The same view of empty stainless commercial kitchen, clean surfaces, overhead lighting, wide angle, with a foreign object composited over 11.08% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S kitchen clean: object removed

Empty stainless commercial kitchen, clean surfaces, overhead lighting, wide angle, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 86.716/255 and from the ground-truth plate by 51.166/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 5.917/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S kitchen clean: removal mask

The exact footprint of the object sitting over empty stainless commercial kitchen, clean surfaces, overhead lighting, wide angle, as an 8-bit mask covering 11.08% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S office: clean plate (ground truth)

A clean 512x512 crop of an empty modern open plan office with desks and monitors, late afternoon sun through blinds, no people, taken from the published plate s-office.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S office: object composited in

The same view of an empty modern open plan office with desks and monitors, late afternoon sun through blinds, no people, with a foreign object composited over 4.75% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S office: object removed

An empty modern open plan office with desks and monitors, late afternoon sun through blinds, no people, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 20.037/255 and from the ground-truth plate by 16.793/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.342/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

S office: removal mask

The exact footprint of the object sitting over an empty modern open plan office with desks and monitors, late afternoon sun through blinds, no people, as an 8-bit mask covering 4.75% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp cyclists: clean plate (ground truth)

A clean 512x512 crop of group of road cyclists on a country road, daylight, taken from the published plate sp-cyclists.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp cyclists: object composited in

The same view of group of road cyclists on a country road, daylight, with a foreign object composited over 9.72% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp cyclists: object removed

Group of road cyclists on a country road, daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 44.39/255 and from the ground-truth plate by 13.229/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 6.53/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp cyclists: removal mask

The exact footprint of the object sitting over group of road cyclists on a country road, daylight, as an 8-bit mask covering 9.72% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp football: clean plate (ground truth)

A clean 512x512 crop of amateur football match on a grass pitch, players contesting the ball, daylight, taken from the published plate sp-football.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp football: object composited in

The same view of amateur football match on a grass pitch, players contesting the ball, daylight, with a foreign object composited over 16.40% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp football: object removed

Amateur football match on a grass pitch, players contesting the ball, daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 50.428/255 and from the ground-truth plate by 59.266/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.641/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp football: removal mask

The exact footprint of the object sitting over amateur football match on a grass pitch, players contesting the ball, daylight, as an 8-bit mask covering 16.40% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp swimmer: clean plate (ground truth)

A clean 512x512 crop of swimmer mid stroke in an indoor pool lane, pool lighting, taken from the published plate sp-swimmer.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp swimmer: object composited in

The same view of swimmer mid stroke in an indoor pool lane, pool lighting, with a foreign object composited over 10.91% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp swimmer: object removed

Swimmer mid stroke in an indoor pool lane, pool lighting, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 53.16/255 and from the ground-truth plate by 54.228/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.358/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

Sp swimmer: removal mask

The exact footprint of the object sitting over swimmer mid stroke in an indoor pool lane, pool lighting, as an 8-bit mask covering 10.91% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad cool: clean plate (ground truth)

A clean 512x512 crop of smooth abstract gradient, deep teal to midnight blue, subtle noise, taken from the published plate t-grad-cool.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad cool: object composited in

The same view of smooth abstract gradient, deep teal to midnight blue, subtle noise, with a foreign object composited over 16.26% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad cool: object removed

Smooth abstract gradient, deep teal to midnight blue, subtle noise, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 80.562/255 and from the ground-truth plate by 6.576/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 2.895/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad cool: removal mask

The exact footprint of the object sitting over smooth abstract gradient, deep teal to midnight blue, subtle noise, as an 8-bit mask covering 16.26% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad warm: clean plate (ground truth)

A clean 512x512 crop of smooth abstract gradient, warm amber to deep rose, soft film grain, taken from the published plate t-grad-warm.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad warm: object composited in

The same view of smooth abstract gradient, warm amber to deep rose, soft film grain, with a foreign object composited over 12.70% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad warm: object removed

Smooth abstract gradient, warm amber to deep rose, soft film grain, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 59.829/255 and from the ground-truth plate by 7.845/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 9.08/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T grad warm: removal mask

The exact footprint of the object sitting over smooth abstract gradient, warm amber to deep rose, soft film grain, as an 8-bit mask covering 12.70% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T ph soft3: clean plate (ground truth)

A clean 512x512 crop of soft blurred abstract background, muted sage green, very low detail, taken from the published plate t-ph-soft3.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T ph soft3: object composited in

The same view of soft blurred abstract background, muted sage green, very low detail, with a foreign object composited over 3.85% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T ph soft3: object removed

Soft blurred abstract background, muted sage green, very low detail, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 41.655/255 and from the ground-truth plate by 3.338/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 2.982/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T ph soft3: removal mask

The exact footprint of the object sitting over soft blurred abstract background, muted sage green, very low detail, as an 8-bit mask covering 3.85% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T tex denim: clean plate (ground truth)

A clean 512x512 crop of close up indigo denim fabric weave, taken from the published plate t-tex-denim.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T tex denim: object composited in

The same view of close up indigo denim fabric weave, with a foreign object composited over 7.10% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T tex denim: object removed

Close up indigo denim fabric weave, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 61.565/255 and from the ground-truth plate by 12.504/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 10.632/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

T tex denim: removal mask

The exact footprint of the object sitting over close up indigo denim fabric weave, as an 8-bit mask covering 7.10% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V bike city: clean plate (ground truth)

A clean 512x512 crop of city bicycle leaning against a brick wall, soft daylight, taken from the published plate v-bike-city.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V bike city: object composited in

The same view of city bicycle leaning against a brick wall, soft daylight, with a foreign object composited over 7.55% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V bike city: object removed

City bicycle leaning against a brick wall, soft daylight, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 53.072/255 and from the ground-truth plate by 26.044/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 7.882/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V bike city: removal mask

The exact footprint of the object sitting over city bicycle leaning against a brick wall, soft daylight, as an 8-bit mask covering 7.55% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V boat harbour: clean plate (ground truth)

A clean 512x512 crop of small fishing boat moored in a harbour, morning light, taken from the published plate v-boat-harbour.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V boat harbour: object composited in

The same view of small fishing boat moored in a harbour, morning light, with a foreign object composited over 5.61% of the frame as a single organic region, neither a rectangle nor an ellipse. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V boat harbour: object removed

Small fishing boat moored in a harbour, morning light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 30.711/255 and from the ground-truth plate by 25.736/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 7.026/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V boat harbour: removal mask

The exact footprint of the object sitting over small fishing boat moored in a harbour, morning light, as an 8-bit mask covering 5.61% of the frame as a single organic region, neither a rectangle nor an ellipse. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V car ev: clean plate (ground truth)

A clean 512x512 crop of white electric hatchback at a charging point, urban car park, taken from the published plate v-car-ev.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V car ev: object composited in

The same view of white electric hatchback at a charging point, urban car park, with a foreign object composited over 4.31% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V car ev: object removed

White electric hatchback at a charging point, urban car park, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 89.979/255 and from the ground-truth plate by 38.547/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 4.784/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V car ev: removal mask

The exact footprint of the object sitting over white electric hatchback at a charging point, urban car park, as an 8-bit mask covering 4.31% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V forklift: clean plate (ground truth)

A clean 512x512 crop of yellow forklift in a warehouse aisle, industrial lighting, taken from the published plate v-forklift.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V forklift: object composited in

The same view of yellow forklift in a warehouse aisle, industrial lighting, with a foreign object composited over 6.26% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V forklift: object removed

Yellow forklift in a warehouse aisle, industrial lighting, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 54.408/255 and from the ground-truth plate by 33.846/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 7.135/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V forklift: removal mask

The exact footprint of the object sitting over yellow forklift in a warehouse aisle, industrial lighting, as an 8-bit mask covering 6.26% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V plane gate: clean plate (ground truth)

A clean 512x512 crop of passenger aircraft at an airport gate, overcast sky, taken from the published plate v-plane-gate.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V plane gate: object composited in

The same view of passenger aircraft at an airport gate, overcast sky, with a foreign object composited over 14.90% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V plane gate: object removed

Passenger aircraft at an airport gate, overcast sky, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 84.008/255 and from the ground-truth plate by 64.23/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 7.224/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V plane gate: removal mask

The exact footprint of the object sitting over passenger aircraft at an airport gate, overcast sky, as an 8-bit mask covering 14.90% of the frame as two disconnected regions - the case a reader that keeps only the largest connected component, or takes the bounding box of both, gets wrong. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V truck artic: clean plate (ground truth)

A clean 512x512 crop of articulated lorry at a distribution depot loading bay, morning light, taken from the published plate v-truck-artic.png before anything was added to it. This is the ANSWER KEY for its group: the object in the source file was composited onto this image, so this is exactly what was behind it. Nothing else in the group came from a second tool's guess.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V truck artic: object composited in

The same view of articulated lorry at a distribution depot loading bay, morning light, with a foreign object composited over 8.83% of the frame as one long thin region, where growing the mask by 12 pixels more than doubles the area it covers. This is the file an object-removal tool is given. Outside the mask it is byte-identical to the clean plate beside it, so any difference a tool leaves there is damage it did rather than content it was handed.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V truck artic: object removed

Articulated lorry at a distribution depot loading bay, morning light, with the object taken back out by Stable Diffusion 1.5 inpainting and the gap reconstructed from the surrounding context alone - the masked latents are erased before sampling, so the model never saw what it was painting over. Inside the mask it differs from the source by 55.181/255 and from the ground-truth plate by 63.287/255; the second number is NOT expected to be small, because an inpainter invents plausible content rather than recovering what was there. Beyond a 16-pixel ring around the mask the frame changes by only 6.781/255, which is the full-frame VAE round trip and not an edit.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

V truck artic: removal mask

The exact footprint of the object sitting over articulated lorry at a distribution depot loading bay, morning light, as an 8-bit mask covering 8.83% of the frame as one long thin region, where growing the mask by 12 pixels more than doubles the area it covers. It holds only the values 0 and 255. The footprint is what DREW the object, so it is ground truth by construction rather than a segmentation of it. Hard-edged on purpose: a feathered edge has no exact footprint, and the exactness is the point of shipping it.

File: PNG · Inpaint Cuts · 512 × 512 px
Use case: Inpainting· Conversion set

JPEG Artefacts

A barista in her twenties (JPEG q25)

The same plate encoded once at JPEG quality 25. Blocking and ringing are bounded by that single pass, so a restorer can be scored against the PNG reference in this group rather than against an opinion Plate owner-cafe.

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A barista in her twenties (JPEG q50)

The same plate encoded once at JPEG quality 50. Blocking and ringing are bounded by that single pass, so a restorer can be scored against the PNG reference in this group rather than against an opinion Plate owner-cafe.

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A barista in her twenties (JPEG q75)

The same plate encoded once at JPEG quality 75. Blocking and ringing are bounded by that single pass, so a restorer can be scored against the PNG reference in this group rather than against an opinion Plate owner-cafe.

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A Black man in his seventies (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A Black man in his seventies (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A Black man in his seventies (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A Black woman hairdresser in (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A Black woman hairdresser in (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A Black woman hairdresser in (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A cappuccino with latte art in (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A cappuccino with latte art in (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A cappuccino with latte art in (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A cluttered joinery workshop (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A cluttered joinery workshop (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A cluttered joinery workshop (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A delivery driver with a (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery driver with a (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery driver with a (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery note on a clipboard (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery note on a clipboard (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery note on a clipboard (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A hairdresser in her forties (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A hairdresser in her forties (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A hairdresser in her forties (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A market stall vendor (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A market stall vendor (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A market stall vendor (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A Middle Eastern man chef (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A Middle Eastern man chef (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A Middle Eastern man chef (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A person's hands signing a (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A person's hands signing a (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A person's hands signing a (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A plated main course on a, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

A plus size white woman in her (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A plus size white woman in her (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A plus size white woman in her (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A scatter of crumpled paper (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A scatter of crumpled paper (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A scatter of crumpled paper (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A smiling woman with short (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A smiling woman with short (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A smiling woman with short (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A South Asian woman (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A South Asian woman (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A South Asian woman (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

A warehouse worker holding a (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A warehouse worker holding a (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A warehouse worker holding a (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A wheelchair user working at (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A wheelchair user working at (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

A wheelchair user working at (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Abstract album cover artwork (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Abstract album cover artwork (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Abstract album cover artwork (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, coral (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, coral (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, coral (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, violet (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, violet (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, violet (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract warm bokeh lights out (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract warm bokeh lights out (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract warm bokeh lights out (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Amateur football match on a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Amateur football match on a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Amateur football match on a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

An East Asian man in his (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An East Asian man in his (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An East Asian man in his (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An electrician in work clothes (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

An electrician in work clothes (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

An electrician in work clothes (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

An empty modern open plan (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An empty modern open plan (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An empty modern open plan (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An older man with a white (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

An older man with a white (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

An older man with a white (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

An older woman with a walking (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An older woman with a walking (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An older woman with a walking (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Arcade racing game track (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Arcade racing game track (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Arcade racing game track (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Articulated lorry at a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Articulated lorry at a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Articulated lorry at a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Black woman mechanic in (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Black woman mechanic in (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Black woman mechanic in (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Book cover artwork, empty rain (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, empty rain (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, empty rain (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, misty (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, misty (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, misty (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, weathered (JPEG q25)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, weathered (JPEG q50)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, weathered (JPEG q75)

File: JPG · JPEG Artefacts · 832 × 1216 px
Use case: Media handling· Conversion set

Brown horse standing in a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Brown horse standing in a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Brown horse standing in a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Canvas backpack on a plain (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Canvas backpack on a plain (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Canvas backpack on a plain (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City bicycle leaning against a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City bicycle leaning against a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City bicycle leaning against a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City bus at a bus stop, grey (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City bus at a bus stop, grey (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City bus at a bus stop, grey (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City skyline across a river at (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City skyline across a river at (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

City skyline across a river at (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Climber on an indoor (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Climber on an indoor (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Climber on an indoor (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up brown leather grain (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up brown leather grain (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up brown leather grain (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up indigo denim fabric (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up indigo denim fabric (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up indigo denim fabric (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coarse recycled (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coarse recycled (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coarse recycled (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coffee being (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up of coffee being (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up of coffee being (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up woven linen fabric (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up woven linen fabric (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Close up woven linen fabric (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Colourful stylised platformer (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Colourful stylised platformer (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Colourful stylised platformer (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Commuter train at a station (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Commuter train at a station (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Commuter train at a station (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Concrete spiral stairwell seen (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Concrete spiral stairwell seen (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Concrete spiral stairwell seen (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark dungeon interior video (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark dungeon interior video (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark dungeon interior video (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark moody blurred background, deep (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, deep (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, deep (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, near (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, near (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, near (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Dark saloon car on a wet city (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark saloon car on a wet city (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark saloon car on a wet city (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

East african woman cleaner (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

East african woman cleaner (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

East african woman cleaner (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Electric scooter parked on a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Electric scooter parked on a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Electric scooter parked on a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty cafe interior before (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty cafe interior before (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty cafe interior before (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty open plan office with (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty open plan office with (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty open plan office with (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty outdoor market stall (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty outdoor market stall (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty outdoor market stall (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty stainless commercial (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty stainless commercial (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty stainless commercial (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Filipino woman nurse at a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Filipino woman nurse at a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Filipino woman nurse at a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Five colleagues of different (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Five colleagues of different (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Five colleagues of different (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Flat concrete wall texture (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat concrete wall texture (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat concrete wall texture (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat dark slate stone surface (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat dark slate stone surface (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat dark slate stone surface (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat fine sand surface with (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat fine sand surface with (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat fine sand surface with (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat lay of textured handmade (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat lay of textured handmade (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat lay of textured handmade (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat oak wood grain surface (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat oak wood grain surface (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat oak wood grain surface (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat red brick wall (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat red brick wall (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat red brick wall (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat rusted steel plate (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat rusted steel plate (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat rusted steel plate (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat white marble slab with (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat white marble slab with (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flat white marble slab with (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Flock of sheep on a green (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Flock of sheep on a green (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Flock of sheep on a green (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Four colleagues around a (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Four colleagues around a (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Four colleagues around a (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Glass water bottle on a plain (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Glass water bottle on a plain (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Glass water bottle on a plain (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Golden retriever sitting on (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Golden retriever sitting on (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Golden retriever sitting on (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Green farm tractor in a field (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Green farm tractor in a field (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Green farm tractor in a field (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Group of road cyclists on a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Group of road cyclists on a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Group of road cyclists on a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a bearded (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a bearded (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a bearded (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a middle (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a middle (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a middle (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a south (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a south (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a south (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

40.5 KB

Head and shoulders portrait of a, plate C (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

62.8 KB

Head and shoulders portrait of a, plate C (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

99.3 KB

Head and shoulders portrait of a, plate C (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

33.8 KB

Head and shoulders portrait of a, plate D (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

46.9 KB

Head and shoulders portrait of a, plate D (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

70.7 KB

Head and shoulders portrait of a, plate D (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

30.4 KB

Head and shoulders portrait of a, plate E (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

45.1 KB

Head and shoulders portrait of a, plate E (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

68.9 KB

Head and shoulders portrait of a, plate E (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an east (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an east (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an east (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

41.3 KB

Head and shoulders portrait of an, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

61.5 KB

Head and shoulders portrait of an, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

jpg

92.7 KB

Head and shoulders portrait of an, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Isometric stylised village (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Isometric stylised village (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Isometric stylised village (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

JPEG Artefact Clean Reference (512px)

Lossless PNG ground truth for JPEG blocking / ringing artefact repair tests.

File: PNG · JPEG Artefacts · 512 × 512 px
Use case: Image enhancementCompression testing+1· Conversion set

JPEG Artefacts: Quality 10 (512px)

Fruit still life saved at JPEG quality 10: heavy blocking/ringing for artefact-removal and enhance tools.

File: JPG · JPEG Artefacts · 512 × 512 px
Use case: Image enhancementCompression testing+1· Conversion set

JPEG Artefacts: Quality 15 (512px)

Fruit still life saved at JPEG quality 15: heavy blocking/ringing for artefact-removal and enhance tools.

File: JPG · JPEG Artefacts · 512 × 512 px
Use case: Image enhancementCompression testing+1· Conversion set

JPEG Artefacts: Quality 25 (512px)

Fruit still life saved at JPEG quality 25: heavy blocking/ringing for artefact-removal and enhance tools.

File: JPG · JPEG Artefacts · 512 × 512 px
Use case: Image enhancementCompression testing+1· Conversion set

JPEG Artefacts: Quality 40 (512px)

Fruit still life saved at JPEG quality 40: heavy blocking/ringing for artefact-removal and enhance tools.

File: JPG · JPEG Artefacts · 512 × 512 px
Use case: Image enhancementCompression testing+1· Conversion set

JPEG Artefacts: Quality 5 (512px)

Fruit still life saved at JPEG quality 5: heavy blocking/ringing for artefact-removal and enhance tools.

File: JPG · JPEG Artefacts · 512 × 512 px
Use case: Image enhancementCompression testing+1· Conversion set

Latina teacher beside a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Latina teacher beside a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Latina teacher beside a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Low poly stylised island with (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Low poly stylised island with (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Low poly stylised island with (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged indian man (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged indian man (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged indian man (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged white male butcher (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged white male butcher (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged white male butcher (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle eastern man tailor (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle eastern man tailor (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle eastern man tailor (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Modern minimalist living room (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Modern minimalist living room (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Modern minimalist living room (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Moody album cover artwork (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Moody album cover artwork (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Moody album cover artwork (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Mountain valley with a river (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Mountain valley with a river (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Mountain valley with a river (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Mud splattered pickup truck on (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Mud splattered pickup truck on (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Mud splattered pickup truck on (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Narrow european old town (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Narrow european old town (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Narrow european old town (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Older white man farmer (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Older white man farmer (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Older white man farmer (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Open laptop on a plain white (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Open laptop on a plain white (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Open laptop on a plain white (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Over ear headphones on a plain (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Over ear headphones on a plain (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Over ear headphones on a plain (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead colourful garden (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead colourful garden (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead colourful garden (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Passenger aircraft at an (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Passenger aircraft at an (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Passenger aircraft at an (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Path through a deciduous (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Path through a deciduous (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Path through a deciduous (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Person lifting weights in a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Person lifting weights in a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Person lifting weights in a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Pixel art side scrolling game (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Pixel art side scrolling game (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Pixel art side scrolling game (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Potted houseplant on a plain (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Potted houseplant on a plain (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Potted houseplant on a plain (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Red vintage convertible parked (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Red vintage convertible parked (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Red vintage convertible parked (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Retro album cover artwork (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Retro album cover artwork (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Retro album cover artwork (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Rocky coastline with breaking (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Rocky coastline with breaking (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Rocky coastline with breaking (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Runner on a park path in the (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Runner on a park path in the (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Runner on a park path in the (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Rustic sourdough loaves on a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Rustic sourdough loaves on a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Rustic sourdough loaves on a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Science fiction video game (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Science fiction video game (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Science fiction video game (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Senior south asian woman (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Senior south asian woman (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Senior south asian woman (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Single running shoe on a plain (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Single running shoe on a plain (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Single running shoe on a plain (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small fishing boat moored in a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small fishing boat moored in a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small fishing boat moored in a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small garden birds on a wooden (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small garden birds on a wooden (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small garden birds on a wooden (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small independent grocery (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small independent grocery (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Small independent grocery (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Smooth abstract gradient (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, warm (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, warm (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, warm (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth vertical gradient from (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth vertical gradient from (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth vertical gradient from (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, pale (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, pale (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, pale (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, warm (JPEG q25)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, warm (JPEG q50)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, warm (JPEG q75)

File: JPG · JPEG Artefacts · 1280 × 853 px
Use case: Media handling· Conversion set

Sports motorcycle parked at a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Sports motorcycle parked at a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Sports motorcycle parked at a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Steel truss bridge over a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Steel truss bridge over a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Steel truss bridge over a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy knight (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy knight (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy knight (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy mage (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy mage (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy mage (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised science fiction (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised science fiction (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised science fiction (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Swimmer mid stroke in an (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Swimmer mid stroke in an (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Swimmer mid stroke in an (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Tabby cat sitting on a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Tabby cat sitting on a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Tabby cat sitting on a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

The exterior of a small, plate A (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate A (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate A (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate B (JPEG q25)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate B (JPEG q50)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate B (JPEG q75)

File: JPG · JPEG Artefacts · 1216 × 832 px
Use case: Media handling· Conversion set

Tropical fish in a planted (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Tropical fish in a planted (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Tropical fish in a planted (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Voxel art city block, blocky (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Voxel art city block, blocky (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Voxel art city block, blocky (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Warehouse aisle between tall (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Warehouse aisle between tall (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Warehouse aisle between tall (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Wheelchair user chef at a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Wheelchair user chef at a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Wheelchair user chef at a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

White electric hatchback at a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

White electric hatchback at a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

White electric hatchback at a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

White panel delivery van (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

White panel delivery van (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

White panel delivery van (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Wristwatch on a plain grey (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Wristwatch on a plain grey (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Wristwatch on a plain grey (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Yellow forklift in a warehouse (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Yellow forklift in a warehouse (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Yellow forklift in a warehouse (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Young east asian woman baker (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Young east asian woman baker (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Young east asian woman baker (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Young man barista signing to a (JPEG q25)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Young man barista signing to a (JPEG q50)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Young man barista signing to a (JPEG q75)

File: JPG · JPEG Artefacts · 1024 × 1024 px
Use case: Media handling· Conversion set

Matting Set

Matting: Fine Fringe Clean RGB (288px)

Clean RGB composite of the fine-fringe soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Fine Fringe Soft Alpha (288px)

Soft-alpha cut-out (hair edges: fine-fringe) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Matting: Fur Edge Clean RGB (288px)

Clean RGB composite of the fur-edge soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Fur Edge Soft Alpha (288px)

Soft-alpha cut-out (hair edges: fur-edge) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Matting: Glass Rim Clean RGB (288px)

Clean RGB composite of the glass-rim soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Glass Rim Soft Alpha (288px)

Soft-alpha cut-out (soft edges: glass-rim) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Matting: Hair Strands Clean RGB (288px)

Clean RGB composite of the hair-strands soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Hair Strands Soft Alpha (288px)

Soft-alpha cut-out (hair edges: hair-strands) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Matting: Plant Edge Clean RGB (288px)

Clean RGB composite of the plant-edge soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Plant Edge Soft Alpha (288px)

Soft-alpha cut-out (soft edges: plant-edge) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Matting: Soft Edge Blob Clean RGB (288px)

Clean RGB composite of the soft-edge-blob soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Soft Edge Blob Soft Alpha (288px)

Soft-alpha cut-out (soft edges: soft-edge-blob) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Matting: Wispy Top Clean RGB (288px)

Clean RGB composite of the wispy-top soft-alpha subject on a neutral grey, paired input for matte extractors.

File: PNG · Matting Set · 288 × 288 px
Use case: Image matting· Paired fixture

Matting: Wispy Top Soft Alpha (288px)

Soft-alpha cut-out (hair edges: wispy-top) for alpha-matting / background-removal quality tests.

File: PNG · Matting Set · 288 × 288 px
Use case: Image mattingComputer vision· Paired fixture

Multipage Sprite

png

534 B

Favicon Pack: 16×16 PNG

Synthetic 16×16 favicon-size mark (not a real brand) for favicon pipeline / resize tests.

File: PNG · Multipage Sprite · 16 × 16 px
Use case: Conversion testing· Conversion set

png

848 B

Favicon Pack: 32×32 PNG

Synthetic 32×32 favicon-size mark (not a real brand) for favicon pipeline / resize tests.

File: PNG · Multipage Sprite · 32 × 32 px
Use case: Conversion testing· Conversion set

png

1.2 KB

Favicon Pack: 48×48 PNG

Synthetic 48×48 favicon-size mark (not a real brand) for favicon pipeline / resize tests.

File: PNG · Multipage Sprite · 48 × 48 px
Use case: Conversion testing· Conversion set

png

300 B

Favicon Pack: 64×64 PNG

Synthetic 64×64 favicon-size mark (not a real brand) for favicon pipeline / resize tests.

File: PNG · Multipage Sprite · 64 × 64 px
Use case: Conversion testing· Conversion set

ico

2.6 KB

Favicon Pack: Multi-size ICO

Windows ICO containing 16/32/48 favicon sizes of the synthetic mark, for ICO parsers and extractors.

File: ICO · Multipage Sprite
Use case: Conversion testing· Conversion set

Multipage Scene TIFF (3 pages)

Three-page TIFF of synthetic scenes: fixture for multipage TIFF loaders and page extractors.

File: TIFF · Multipage Sprite · 288 × 288 px
Use case: Conversion testingComputer vision· Conversion set

Single-Page Scene TIFF

Single-page TIFF of the same scene as page 1 of the multipage twin.

File: TIFF · Multipage Sprite · 288 × 288 px
Use case: Conversion testing· Conversion set

png

1.6 KB

Sprite Sheet: grid-2x2 (2×2)

RGBA sprite sheet (2×2 cells of 64px) with numbered frames for atlas / CSS-sprite tools.

File: PNG · Multipage Sprite · 2 rows
Use case: Computer visionConversion testing· Conversion set

png

2.5 KB

Sprite Sheet: grid-3x2 (3×2)

RGBA sprite sheet (3×2 cells of 64px) with numbered frames for atlas / CSS-sprite tools.

File: PNG · Multipage Sprite · 2 rows
Use case: Computer visionConversion testing· Conversion set

png

1.5 KB

Sprite Sheet: strip-4 (4×1)

RGBA sprite sheet (4×1 cells of 64px) with numbered frames for atlas / CSS-sprite tools.

File: PNG · Multipage Sprite · 1 row
Use case: Computer visionConversion testing· Conversion set

Two-Page Scene TIFF

Two-page TIFF companion for multipage loader edge cases.

File: TIFF · Multipage Sprite · 288 × 288 px
Use case: Conversion testing· Conversion set

OCR

Business Card (OCR fixture)

A rendered business card with a name, title, company, email, phone, and address, paired with a matching vCard so you can measure OCR and contact-extraction accuracy against a known ground truth.

File: PNG · OCR · 900 × 500 px
Use case: OCR testingContact parsing· Paired fixture

Business Card: vCard (.vcf)

The vCard (.vcf) ground truth for the business-card image: the same name, title, company, email, phone, and address in structured form. The positive reference for testing OCR-to-contact extraction.

File: VCF · OCR
Use case: Contact parsingOCR testing· Paired fixture

delivery (gaussian blur)

A delivery typeset over a synthetic photographic background, gaussian blur. Softened by a recorded Gaussian radius, which is what an out-of-focus capture looks like. 29 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

delivery (low contrast)

A delivery typeset over a synthetic photographic background, low contrast. Gain and offset applied so the text sits close to the ground, which is where thresholding fails. 29 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

delivery (rotated -7 degrees)

A delivery typeset over a synthetic photographic background, rotated -7 degrees. Rotated, with every word box rotated with it, so skew correction can be scored rather than eyeballed. 29 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

delivery (rotated +3 degrees)

A delivery typeset over a synthetic photographic background, rotated +3 degrees. A slight skew of the kind a phone photo of a docket actually has. 29 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

delivery (sensor noise)

A delivery typeset over a synthetic photographic background, sensor noise. Additive noise at a recorded sigma, as a dim-light capture would carry. 29 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

delivery (undistorted)

A delivery typeset over a synthetic photographic background, undistorted. Straight-on and undamaged, the baseline every other variant in this group is measured against. 29 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

invoice (gaussian blur)

A invoice typeset over a synthetic photographic background, gaussian blur. Softened by a recorded Gaussian radius, which is what an out-of-focus capture looks like. 33 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

invoice (low contrast)

A invoice typeset over a synthetic photographic background, low contrast. Gain and offset applied so the text sits close to the ground, which is where thresholding fails. 33 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

invoice (rotated -7 degrees)

A invoice typeset over a synthetic photographic background, rotated -7 degrees. Rotated, with every word box rotated with it, so skew correction can be scored rather than eyeballed. 33 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

invoice (rotated +3 degrees)

A invoice typeset over a synthetic photographic background, rotated +3 degrees. A slight skew of the kind a phone photo of a docket actually has. 33 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

invoice (sensor noise)

A invoice typeset over a synthetic photographic background, sensor noise. Additive noise at a recorded sigma, as a dim-light capture would carry. 33 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

invoice (undistorted)

A invoice typeset over a synthetic photographic background, undistorted. Straight-on and undamaged, the baseline every other variant in this group is measured against. 33 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

receipt (gaussian blur)

A receipt typeset over a synthetic photographic background, gaussian blur. Softened by a recorded Gaussian radius, which is what an out-of-focus capture looks like. 23 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

receipt (low contrast)

A receipt typeset over a synthetic photographic background, low contrast. Gain and offset applied so the text sits close to the ground, which is where thresholding fails. 23 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

receipt (rotated -7 degrees)

A receipt typeset over a synthetic photographic background, rotated -7 degrees. Rotated, with every word box rotated with it, so skew correction can be scored rather than eyeballed. 23 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

receipt (rotated +3 degrees)

A receipt typeset over a synthetic photographic background, rotated +3 degrees. A slight skew of the kind a phone photo of a docket actually has. 23 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

receipt (sensor noise)

A receipt typeset over a synthetic photographic background, sensor noise. Additive noise at a recorded sigma, as a dim-light capture would carry. 23 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

receipt (undistorted)

A receipt typeset over a synthetic photographic background, undistorted. Straight-on and undamaged, the baseline every other variant in this group is measured against. 23 words, each with an exact bounding box in the companion JSON.

File: PNG · OCR
Use case: OCR testing· Conversion set

OCR Hard

OCR Hard GT: Dotted Code

Exact ground-truth string for the dotted-code OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Dotted Serial

Exact ground-truth string for the dotted-serial OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Low Contrast

Exact ground-truth string for the low-contrast OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Low Label

Exact ground-truth string for the low-label OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Low Receipt

Exact ground-truth string for the low-receipt OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Skew Address

Exact ground-truth string for the skew-address OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Skew Date

Exact ground-truth string for the skew-date OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard GT: Skew Invoice

Exact ground-truth string for the skew-invoice OCR hard case.

File: TXT · OCR Hard
Use case: OCR testing· Paired fixture

OCR Hard: Dotted Code

Hard OCR case (dotted): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Dotted Serial

Hard OCR case (dotted): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Low Contrast

Hard OCR case (low-contrast): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Low Label

Hard OCR case (low-contrast): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Low Receipt

Hard OCR case (low-contrast): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Skew Address

Hard OCR case (skew): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Skew Date

Hard OCR case (skew): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

OCR Hard: Skew Invoice

Hard OCR case (skew): SAMPLE text rendered with challenging conditions. Pair with the .txt ground truth.

File: PNG · OCR Hard
Use case: OCR testingComputer vision· Paired fixture

Outpaint Borders

A av f4: expanded canvas

The same view of head and shoulders portrait of an older white woman with short grey hair, plain cream background, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.231/255, the VAE round trip, but over the whole rectangle by 6.293/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

A av f4: frame before expansion

A 512x512 view of head and shoulders portrait of an older white woman with short grey hair, plain cream background - the frame an outpainting tool is given, cropped from the published plate nss-a-av-f4_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

A av f5: expanded canvas

The same view of head and shoulders portrait of an east asian woman with glasses, plain warm grey background, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.482/255, the VAE round trip, but over the whole rectangle by 6.4/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

A av f5: frame before expansion

A 512x512 view of head and shoulders portrait of an east asian woman with glasses, plain warm grey background - the frame an outpainting tool is given, cropped from the published plate nss-a-av-f5_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

A av m3: expanded canvas

The same view of head and shoulders portrait of a bearded white man in his thirties, plain olive studio background, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.277/255, the VAE round trip, but over the whole rectangle by 6.19/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

A av m3: frame before expansion

A 512x512 view of head and shoulders portrait of a bearded white man in his thirties, plain olive studio background - the frame an outpainting tool is given, cropped from the published plate nss-a-av-m3_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

A av m4: expanded canvas

The same view of head and shoulders portrait of a south asian man in his fifties, plain slate studio background, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.007/255, the VAE round trip, but over the whole rectangle by 5.68/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

A av m4: frame before expansion

A 512x512 view of head and shoulders portrait of a south asian man in his fifties, plain slate studio background - the frame an outpainting tool is given, cropped from the published plate nss-a-av-m4_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

png

481 KB

A avatar f1: expanded canvas

The same view of head and shoulders portrait of a smiling white woman against a plain blue studio background, expanded to 512x704 - 192 pixels added below alone, the one case where the origin does NOT move - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.251/255, the VAE round trip, but over the whole rectangle by 4.956/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

png

393.9 KB

A avatar f1: frame before expansion

A 512x512 view of head and shoulders portrait of a smiling white woman against a plain blue studio background - the frame an outpainting tool is given, cropped from the published plate nss-a-avatar-f1_00001_.png. Its companion expands it to 512x704 with 192 pixels added below alone, the one case where the origin does NOT move, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

png

652.5 KB

A avatar f2: expanded canvas

The same view of head and shoulders portrait of a young hijabi woman against a plain neutral studio background, expanded to 704x704 - 96 pixels added on all four sides - so 47.1% of this frame is invented canvas. The original sits at pixels 96,96 to 608,608. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.267/255, the VAE round trip, but over the whole rectangle by 9.706/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 704 px
Use case: Inpainting· Conversion set

png

418.6 KB

A avatar f2: frame before expansion

A 512x512 view of head and shoulders portrait of a young hijabi woman against a plain neutral studio background - the frame an outpainting tool is given, cropped from the published plate nss-a-avatar-f2_00001_.png. Its companion expands it to 704x704 with 96 pixels added on all four sides, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

png

637.8 KB

A avatar m2: expanded canvas

The same view of head and shoulders portrait of an older east asian man against a plain warm studio background, expanded to 768x640 - a different amount on every side, which is the case a tool assuming symmetry fails - so 46.7% of this frame is invented canvas. The original sits at pixels 64,32 to 576,544. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.021/255, the VAE round trip, but over the whole rectangle by 8.184/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 640 px
Use case: Inpainting· Conversion set

png

395.8 KB

A avatar m2: frame before expansion

A 512x512 view of head and shoulders portrait of an older east asian man against a plain warm studio background - the frame an outpainting tool is given, cropped from the published plate nss-a-avatar-m2_00001_.png. Its companion expands it to 768x640 with a different amount on every side, which is the case a tool assuming symmetry fails, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

C album abs: expanded canvas

The same view of abstract album cover artwork, bold overlapping colour shapes, grainy print texture, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.331/255, the VAE round trip, but over the whole rectangle by 10.141/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

C album abs: frame before expansion

A 512x512 view of abstract album cover artwork, bold overlapping colour shapes, grainy print texture - the frame an outpainting tool is given, cropped from the published plate nss-c-album-abs_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

C album retro: expanded canvas

The same view of retro album cover artwork, sunset gradient and geometric lines, 1980s print feel, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 9.476/255, the VAE round trip, but over the whole rectangle by 10.985/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

C album retro: frame before expansion

A 512x512 view of retro album cover artwork, sunset gradient and geometric lines, 1980s print feel - the frame an outpainting tool is given, cropped from the published plate nss-c-album-retro_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

C book hist: expanded canvas

The same view of book cover artwork, weathered antique map and brass instruments on wood, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 7.793/255, the VAE round trip, but over the whole rectangle by 8.625/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

C book hist: frame before expansion

A 512x512 view of book cover artwork, weathered antique map and brass instruments on wood - the frame an outpainting tool is given, cropped from the published plate nss-c-book-hist_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

C book nature: expanded canvas

The same view of book cover artwork, misty mountain ridge at dawn, layered depth, serene, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 2.563/255, the VAE round trip, but over the whole rectangle by 3.277/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

C book nature: frame before expansion

A 512x512 view of book cover artwork, misty mountain ridge at dawn, layered depth, serene - the frame an outpainting tool is given, cropped from the published plate nss-c-book-nature_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

D clipboard: expanded canvas

The same view of a delivery note on a clipboard held against a van door, outdoor overcast light, slight angle, expanded to 512x704 - 192 pixels added above alone, so every row index shifts - so 27.3% of this frame is invented canvas. The original sits at pixels 0,192 to 512,704. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.301/255, the VAE round trip, but over the whole rectangle by 6.294/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

D clipboard: frame before expansion

A 512x512 view of a delivery note on a clipboard held against a van door, outdoor overcast light, slight angle - the frame an outpainting tool is given, cropped from the published plate nss-d-clipboard_00001_.png. Its companion expands it to 512x704 with 192 pixels added above alone, so every row index shifts, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

D invoice: expanded canvas

The same view of a printed invoice on a wooden desk beside a calculator and a pen, flat lay, soft daylight, expanded to 512x704 - 192 pixels added below alone, the one case where the origin does NOT move - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.626/255, the VAE round trip, but over the whole rectangle by 5.717/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

D invoice: frame before expansion

A 512x512 view of a printed invoice on a wooden desk beside a calculator and a pen, flat lay, soft daylight - the frame an outpainting tool is given, cropped from the published plate nss-d-invoice_00001_.png. Its companion expands it to 512x704 with 192 pixels added below alone, the one case where the origin does NOT move, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

D menu: expanded canvas

The same view of a restaurant menu card standing on a laid table, cutlery and glassware, warm evening light, expanded to 704x704 - 96 pixels added on all four sides - so 47.1% of this frame is invented canvas. The original sits at pixels 96,96 to 608,608. Compare it INSET by 40 pixels: over that core it differs from the original by only 7.121/255, the VAE round trip, but over the whole rectangle by 12.793/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 704 px
Use case: Inpainting· Conversion set

D menu: frame before expansion

A 512x512 view of a restaurant menu card standing on a laid table, cutlery and glassware, warm evening light - the frame an outpainting tool is given, cropped from the published plate nss-d-menu_00001_.png. Its companion expands it to 704x704 with 96 pixels added on all four sides, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

D receipts: expanded canvas

The same view of a scatter of crumpled paper receipts on a kitchen table, overhead view, harsh afternoon sun and hard shadows, expanded to 768x640 - a different amount on every side, which is the case a tool assuming symmetry fails - so 46.7% of this frame is invented canvas. The original sits at pixels 64,32 to 576,544. Compare it INSET by 40 pixels: over that core it differs from the original by only 8.885/255, the VAE round trip, but over the whole rectangle by 15.578/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 640 px
Use case: Inpainting· Conversion set

D receipts: frame before expansion

A 512x512 view of a scatter of crumpled paper receipts on a kitchen table, overhead view, harsh afternoon sun and hard shadows - the frame an outpainting tool is given, cropped from the published plate nss-d-receipts_00001_.png. Its companion expands it to 768x640 with a different amount on every side, which is the case a tool assuming symmetry fails, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

F bread rustic: expanded canvas

The same view of rustic sourdough loaves on a wooden board, flour dust, warm bakery light, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 10.419/255, the VAE round trip, but over the whole rectangle by 12.644/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

F bread rustic: frame before expansion

A 512x512 view of rustic sourdough loaves on a wooden board, flour dust, warm bakery light - the frame an outpainting tool is given, cropped from the published plate nss-f-bread-rustic_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

F coffee pour: expanded canvas

The same view of close up of coffee being poured into a white cup, steam, cafe counter, shallow depth, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.288/255, the VAE round trip, but over the whole rectangle by 5.818/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

F coffee pour: frame before expansion

A 512x512 view of close up of coffee being poured into a white cup, steam, cafe counter, shallow depth - the frame an outpainting tool is given, cropped from the published plate nss-f-coffee-pour_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

F plate pasta: expanded canvas

The same view of overhead plate of fresh pasta with herbs on a linen cloth, natural window light, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 8.016/255, the VAE round trip, but over the whole rectangle by 10.024/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

F plate pasta: frame before expansion

A 512x512 view of overhead plate of fresh pasta with herbs on a linen cloth, natural window light - the frame an outpainting tool is given, cropped from the published plate nss-f-plate-pasta_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

F plate salad: expanded canvas

The same view of overhead colourful garden salad in a ceramic bowl, marble surface, soft daylight, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 8.419/255, the VAE round trip, but over the whole rectangle by 11.066/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

F plate salad: frame before expansion

A 512x512 view of overhead colourful garden salad in a ceramic bowl, marble surface, soft daylight - the frame an outpainting tool is given, cropped from the published plate nss-f-plate-salad_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

G char mage: expanded canvas

The same view of stylised fantasy mage character render, full body, neutral grey backdrop, expanded to 512x704 - 192 pixels added above alone, so every row index shifts - so 27.3% of this frame is invented canvas. The original sits at pixels 0,192 to 512,704. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.477/255, the VAE round trip, but over the whole rectangle by 6.674/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

G char mage: frame before expansion

A 512x512 view of stylised fantasy mage character render, full body, neutral grey backdrop - the frame an outpainting tool is given, cropped from the published plate nss-g-char-mage_00001_.png. Its companion expands it to 512x704 with 192 pixels added above alone, so every row index shifts, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

G fantasy keep: expanded canvas

The same view of fantasy video game environment, stone keep on a cliff at sunset, expanded to 704x704 - 96 pixels added on all four sides - so 47.1% of this frame is invented canvas. The original sits at pixels 96,96 to 608,608. Compare it INSET by 40 pixels: over that core it differs from the original by only 7.895/255, the VAE round trip, but over the whole rectangle by 10.907/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 704 px
Use case: Inpainting· Conversion set

G fantasy keep: frame before expansion

A 512x512 view of fantasy video game environment, stone keep on a cliff at sunset - the frame an outpainting tool is given, cropped from the published plate nss-g-fantasy-keep_00001_.png. Its companion expands it to 704x704 with 96 pixels added on all four sides, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

G isometric: expanded canvas

The same view of isometric stylised village diorama, soft shading, mobile style, expanded to 768x640 - a different amount on every side, which is the case a tool assuming symmetry fails - so 46.7% of this frame is invented canvas. The original sits at pixels 64,32 to 576,544. Compare it INSET by 40 pixels: over that core it differs from the original by only 8.719/255, the VAE round trip, but over the whole rectangle by 12.374/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 640 px
Use case: Inpainting· Conversion set

G isometric: frame before expansion

A 512x512 view of isometric stylised village diorama, soft shading, mobile style - the frame an outpainting tool is given, cropped from the published plate nss-g-isometric_00001_.png. Its companion expands it to 768x640 with a different amount on every side, which is the case a tool assuming symmetry fails, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

G platformer: expanded canvas

The same view of colourful stylised platformer level, floating islands, bright palette, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 7.477/255, the VAE round trip, but over the whole rectangle by 8.353/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

G platformer: frame before expansion

A 512x512 view of colourful stylised platformer level, floating islands, bright palette - the frame an outpainting tool is given, cropped from the published plate nss-g-platformer_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Invoice desk: expanded canvas

The same view of a printed invoice on a wooden desk beside a calculator and a pen, flat lay, soft daylight, expanded to 512x704 - 192 pixels added above alone, so every row index shifts - so 27.3% of this frame is invented canvas. The original sits at pixels 0,192 to 512,704. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.206/255, the VAE round trip, but over the whole rectangle by 8.328/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

Invoice desk: frame before expansion

A 512x512 view of a printed invoice on a wooden desk beside a calculator and a pen, flat lay, soft daylight - the frame an outpainting tool is given, cropped from the published plate nss-invoice-desk_00001_.png. Its companion expands it to 512x704 with 192 pixels added above alone, so every row index shifts, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

L forest path: expanded canvas

The same view of path through a deciduous forest, dappled sunlight, receding depth, expanded to 512x704 - 192 pixels added below alone, the one case where the origin does NOT move - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 10.599/255, the VAE round trip, but over the whole rectangle by 11.516/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

L forest path: frame before expansion

A 512x512 view of path through a deciduous forest, dappled sunlight, receding depth - the frame an outpainting tool is given, cropped from the published plate nss-l-forest-path_00001_.png. Its companion expands it to 512x704 with 192 pixels added below alone, the one case where the origin does NOT move, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

L interior mod: expanded canvas

The same view of modern minimalist living room interior, large windows, natural daylight, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 9.931/255, the VAE round trip, but over the whole rectangle by 13.41/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

L interior mod: frame before expansion

A 512x512 view of modern minimalist living room interior, large windows, natural daylight - the frame an outpainting tool is given, cropped from the published plate nss-l-interior-mod_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Ledger book: expanded canvas

The same view of an open paper accounts ledger with ruled columns on a desk, fountain pen resting on the page, window light, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.179/255, the VAE round trip, but over the whole rectangle by 6.829/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

Ledger book: frame before expansion

A 512x512 view of an open paper accounts ledger with ruled columns on a desk, fountain pen resting on the page, window light - the frame an outpainting tool is given, cropped from the published plate nss-ledger-book_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

N birds feed: expanded canvas

The same view of small garden birds on a wooden feeder, natural daylight, expanded to 704x704 - 96 pixels added on all four sides - so 47.1% of this frame is invented canvas. The original sits at pixels 96,96 to 608,608. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.083/255, the VAE round trip, but over the whole rectangle by 6.785/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 704 px
Use case: Inpainting· Conversion set

N birds feed: frame before expansion

A 512x512 view of small garden birds on a wooden feeder, natural daylight - the frame an outpainting tool is given, cropped from the published plate nss-n-birds-feed_00001_.png. Its companion expands it to 704x704 with 96 pixels added on all four sides, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

N cat window: expanded canvas

The same view of tabby cat sitting on a windowsill, natural backlight, indoor, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.614/255, the VAE round trip, but over the whole rectangle by 5.838/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

N cat window: frame before expansion

A 512x512 view of tabby cat sitting on a windowsill, natural backlight, indoor - the frame an outpainting tool is given, cropped from the published plate nss-n-cat-window_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

N horse field: expanded canvas

The same view of brown horse standing in a field, overcast daylight, expanded to 768x640 - a different amount on every side, which is the case a tool assuming symmetry fails - so 46.7% of this frame is invented canvas. The original sits at pixels 64,32 to 576,544. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.811/255, the VAE round trip, but over the whole rectangle by 9.977/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 640 px
Use case: Inpainting· Conversion set

N horse field: frame before expansion

A 512x512 view of brown horse standing in a field, overcast daylight - the frame an outpainting tool is given, cropped from the published plate nss-n-horse-field_00001_.png. Its companion expands it to 768x640 with a different amount on every side, which is the case a tool assuming symmetry fails, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Outpaint Full Reference: 128px Border Suite (512px)

Full-frame plant scene used as the target for outpainting tests with a 128px empty border input.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Full Reference: 192px Border Suite (512px)

Full-frame reference for 192px border outpainting.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Full Reference: 64px Border Suite (512px)

Full-frame plant scene used as the target for outpainting tests with a 64px empty border input.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Full Reference: 96px Border Suite (512px)

Full-frame reference for 96px border outpainting.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Input: 128px Empty Border (512px)

Plant content centred with a 128px empty border: classic outpainting / generative expand fixture.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Input: 192px Empty Border (512px)

Content with 192px empty border for outpainting tools.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Input: 64px Empty Border (512px)

Plant content centred with a 64px empty border: classic outpainting / generative expand fixture.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Input: 96px Empty Border (512px)

Content with 96px empty border for outpainting tools.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Mask: 128px Border (512px)

White = region to outpaint / expand; black = known content.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Mask: 192px Border (512px)

Expand-region mask for outpainting.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Mask: 64px Border (512px)

White = region to outpaint / expand; black = known content.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Outpaint Mask: 96px Border (512px)

Expand-region mask for outpainting.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: OutpaintingImage pipeline QA· Conversion set

Owner market: expanded canvas

The same view of a market stall vendor arranging fresh vegetables on his stall, outdoor daylight, crates and price cards, documentary, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 10.987/255, the VAE round trip, but over the whole rectangle by 12.497/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

Owner market: frame before expansion

A 512x512 view of a market stall vendor arranging fresh vegetables on his stall, outdoor daylight, crates and price cards, documentary - the frame an outpainting tool is given, cropped from the published plate nss-owner-market_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P accountant: expanded canvas

The same view of middle aged indian man accountant at a desk with paper ledgers and a calculator, desk lamp, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 3.731/255, the VAE round trip, but over the whole rectangle by 4.487/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

P accountant: frame before expansion

A 512x512 view of middle aged indian man accountant at a desk with paper ledgers and a calculator, desk lamp - the frame an outpainting tool is given, cropped from the published plate nss-p-accountant_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P chef wheel: expanded canvas

The same view of wheelchair user chef at a lowered prep counter slicing vegetables, professional kitchen, expanded to 512x704 - 192 pixels added above alone, so every row index shifts - so 27.3% of this frame is invented canvas. The original sits at pixels 0,192 to 512,704. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.544/255, the VAE round trip, but over the whole rectangle by 6.677/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

P chef wheel: frame before expansion

A 512x512 view of wheelchair user chef at a lowered prep counter slicing vegetables, professional kitchen - the frame an outpainting tool is given, cropped from the published plate nss-p-chef-wheel_00001_.png. Its companion expands it to 512x704 with 192 pixels added above alone, so every row index shifts, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P driver eve: expanded canvas

The same view of a delivery driver with a parcel and a handheld scanner at a residential door, early evening light, van blurred behind, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.174/255, the VAE round trip, but over the whole rectangle by 7.379/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

P driver eve: frame before expansion

A 512x512 view of a delivery driver with a parcel and a handheld scanner at a residential door, early evening light, van blurred behind - the frame an outpainting tool is given, cropped from the published plate nss-p-driver-eve_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P farmer field: expanded canvas

The same view of older white man farmer crouching in a vegetable field holding produce, overcast daylight, expanded to 512x704 - 192 pixels added below alone, the one case where the origin does NOT move - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.112/255, the VAE round trip, but over the whole rectangle by 5.098/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

P farmer field: frame before expansion

A 512x512 view of older white man farmer crouching in a vegetable field holding produce, overcast daylight - the frame an outpainting tool is given, cropped from the published plate nss-p-farmer-field_00001_.png. Its companion expands it to 512x704 with 192 pixels added below alone, the one case where the origin does NOT move, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P market sen: expanded canvas

The same view of a Black man in his seventies with grey hair arranging vegetables at an outdoor market stall, overcast daylight, crates and handwritten price cards, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.126/255, the VAE round trip, but over the whole rectangle by 6.273/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

P market sen: frame before expansion

A 512x512 view of a Black man in his seventies with grey hair arranging vegetables at an outdoor market stall, overcast daylight, crates and handwritten price cards - the frame an outpainting tool is given, cropped from the published plate nss-p-market-sen_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P salon mid: expanded canvas

The same view of a Black woman hairdresser in her forties in a bright salon, mirrors and styling chairs behind, professional, looking at camera, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.042/255, the VAE round trip, but over the whole rectangle by 4.466/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

P salon mid: frame before expansion

A 512x512 view of a Black woman hairdresser in her forties in a bright salon, mirrors and styling chairs behind, professional, looking at camera - the frame an outpainting tool is given, cropped from the published plate nss-p-salon-mid_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

P trades w: expanded canvas

The same view of a South Asian woman electrician in work clothes and safety glasses checking a consumer unit, domestic hallway, tool bag open, natural light, expanded to 704x704 - 96 pixels added on all four sides - so 47.1% of this frame is invented canvas. The original sits at pixels 96,96 to 608,608. Compare it INSET by 40 pixels: over that core it differs from the original by only 3.963/255, the VAE round trip, but over the whole rectangle by 10.812/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 704 px
Use case: Inpainting· Conversion set

P trades w: frame before expansion

A 512x512 view of a South Asian woman electrician in work clothes and safety glasses checking a consumer unit, domestic hallway, tool bag open, natural light - the frame an outpainting tool is given, cropped from the published plate nss-p-trades-w_00001_.png. Its companion expands it to 704x704 with 96 pixels added on all four sides, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Pr bottle: expanded canvas

The same view of glass water bottle on a plain white studio background, soft light, expanded to 768x640 - a different amount on every side, which is the case a tool assuming symmetry fails - so 46.7% of this frame is invented canvas. The original sits at pixels 64,32 to 576,544. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.819/255, the VAE round trip, but over the whole rectangle by 7.337/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 640 px
Use case: Inpainting· Conversion set

Pr bottle: frame before expansion

A 512x512 view of glass water bottle on a plain white studio background, soft light - the frame an outpainting tool is given, cropped from the published plate nss-pr-bottle_00001_.png. Its companion expands it to 768x640 with a different amount on every side, which is the case a tool assuming symmetry fails, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Pr plant pot: expanded canvas

The same view of potted houseplant on a plain neutral background, soft daylight, expanded to 512x704 - 192 pixels added above alone, so every row index shifts - so 27.3% of this frame is invented canvas. The original sits at pixels 0,192 to 512,704. Compare it INSET by 40 pixels: over that core it differs from the original by only 3.898/255, the VAE round trip, but over the whole rectangle by 5.136/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

Pr plant pot: frame before expansion

A 512x512 view of potted houseplant on a plain neutral background, soft daylight - the frame an outpainting tool is given, cropped from the published plate nss-pr-plant-pot_00001_.png. Its companion expands it to 512x704 with 192 pixels added above alone, so every row index shifts, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Pr shoe: expanded canvas

The same view of single running shoe on a plain white studio background, soft light, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 9.491/255, the VAE round trip, but over the whole rectangle by 10.432/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

Pr shoe: frame before expansion

A 512x512 view of single running shoe on a plain white studio background, soft light - the frame an outpainting tool is given, cropped from the published plate nss-pr-shoe_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

S bakery: expanded canvas

The same view of the exterior of a small independent bakery on a quiet street, awning and window display, morning light, expanded to 512x704 - 192 pixels added below alone, the one case where the origin does NOT move - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.452/255, the VAE round trip, but over the whole rectangle by 7.26/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

S bakery: frame before expansion

A 512x512 view of the exterior of a small independent bakery on a quiet street, awning and window display, morning light - the frame an outpainting tool is given, cropped from the published plate nss-s-bakery_00001_.png. Its companion expands it to 512x704 with 192 pixels added below alone, the one case where the origin does NOT move, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

S cafe empty: expanded canvas

The same view of empty cafe interior before opening, chairs down, morning sun across tables, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 4.615/255, the VAE round trip, but over the whole rectangle by 8.693/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

S cafe empty: frame before expansion

A 512x512 view of empty cafe interior before opening, chairs down, morning sun across tables - the frame an outpainting tool is given, cropped from the published plate nss-s-cafe-empty_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

S workshop: expanded canvas

The same view of a cluttered joinery workshop bench with hand tools and sawdust, window light from the left, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.862/255, the VAE round trip, but over the whole rectangle by 7.908/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

S workshop: frame before expansion

A 512x512 view of a cluttered joinery workshop bench with hand tools and sawdust, window light from the left - the frame an outpainting tool is given, cropped from the published plate nss-s-workshop_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Sp gym: expanded canvas

The same view of person lifting weights in a gym, indoor lighting, expanded to 704x512 - 192 pixels added to the left alone, so the original's origin moves - so 27.3% of this frame is invented canvas. The original sits at pixels 192,0 to 704,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 2.509/255, the VAE round trip, but over the whole rectangle by 4.2/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

Sp gym: frame before expansion

A 512x512 view of person lifting weights in a gym, indoor lighting - the frame an outpainting tool is given, cropped from the published plate nss-sp-gym_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the left alone, so the original's origin moves, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

T bokeh warm: expanded canvas

The same view of abstract warm bokeh lights out of focus, dark background, expanded to 512x704 - 192 pixels added above alone, so every row index shifts - so 27.3% of this frame is invented canvas. The original sits at pixels 0,192 to 512,704. Compare it INSET by 40 pixels: over that core it differs from the original by only 3.972/255, the VAE round trip, but over the whole rectangle by 5.543/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

T bokeh warm: frame before expansion

A 512x512 view of abstract warm bokeh lights out of focus, dark background - the frame an outpainting tool is given, cropped from the published plate nss-t-bokeh-warm_00001_.png. Its companion expands it to 512x704 with 192 pixels added above alone, so every row index shifts, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

T tex brick: expanded canvas

The same view of flat red brick wall, expanded to 512x704 - 192 pixels added below alone, the one case where the origin does NOT move - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 5.876/255, the VAE round trip, but over the whole rectangle by 7.834/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 704 px
Use case: Inpainting· Conversion set

T tex brick: frame before expansion

A 512x512 view of flat red brick wall - the frame an outpainting tool is given, cropped from the published plate nss-t-tex-brick_00001_.png. Its companion expands it to 512x704 with 192 pixels added below alone, the one case where the origin does NOT move, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

T tex linen: expanded canvas

The same view of close up woven linen fabric texture, natural fibres, expanded to 704x704 - 96 pixels added on all four sides - so 47.1% of this frame is invented canvas. The original sits at pixels 96,96 to 608,608. Compare it INSET by 40 pixels: over that core it differs from the original by only 14.966/255, the VAE round trip, but over the whole rectangle by 16.919/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 704 px
Use case: Inpainting· Conversion set

T tex linen: frame before expansion

A 512x512 view of close up woven linen fabric texture, natural fibres - the frame an outpainting tool is given, cropped from the published plate nss-t-tex-linen_00001_.png. Its companion expands it to 704x704 with 96 pixels added on all four sides, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

V moto sport: expanded canvas

The same view of sports motorcycle parked at a kerb, sunny day, expanded to 768x512 - 128 pixels added to the left and right, the symmetric case - so 33.3% of this frame is invented canvas. The original sits at pixels 128,0 to 640,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 8.192/255, the VAE round trip, but over the whole rectangle by 10.954/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 768 × 512 px
Use case: Inpainting· Conversion set

V moto sport: frame before expansion

A 512x512 view of sports motorcycle parked at a kerb, sunny day - the frame an outpainting tool is given, cropped from the published plate nss-v-moto-sport_00001_.png. Its companion expands it to 768x512 with 128 pixels added to the left and right, the symmetric case, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

V tractor: expanded canvas

The same view of green farm tractor in a field, blue sky, expanded to 512x768 - 128 pixels added above and below, the symmetric case turned on its side - so 33.3% of this frame is invented canvas. The original sits at pixels 0,128 to 512,640. Compare it INSET by 40 pixels: over that core it differs from the original by only 8.434/255, the VAE round trip, but over the whole rectangle by 14.105/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 512 × 768 px
Use case: Inpainting· Conversion set

V tractor: frame before expansion

A 512x512 view of green farm tractor in a field, blue sky - the frame an outpainting tool is given, cropped from the published plate nss-v-tractor_00001_.png. Its companion expands it to 512x768 with 128 pixels added above and below, the symmetric case turned on its side, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Warehouse clip: expanded canvas

The same view of a warehouse worker holding a clipboard beside tall racking, high bay lighting, boxes on pallets, expanded to 704x512 - 192 pixels added to the right alone, so the original is no longer centred - so 27.3% of this frame is invented canvas. The original sits at pixels 0,0 to 512,512. Compare it INSET by 40 pixels: over that core it differs from the original by only 6.222/255, the VAE round trip, but over the whole rectangle by 7.652/255, because the pad deliberately feathers the original's outer edge into the new area. A test that expects the whole rectangle untouched fails on a correct tool.

File: PNG · Outpaint Borders · 704 × 512 px
Use case: Inpainting· Conversion set

Warehouse clip: frame before expansion

A 512x512 view of a warehouse worker holding a clipboard beside tall racking, high bay lighting, boxes on pallets - the frame an outpainting tool is given, cropped from the published plate nss-warehouse-clip_00001_.png. Its companion expands it to 704x512 with 192 pixels added to the right alone, so the original is no longer centred, and the boundary record in this group states the exact rectangle this image occupies inside that frame - so where it ended up can be checked rather than eyeballed.

File: PNG · Outpaint Borders · 512 × 512 px
Use case: Inpainting· Conversion set

Pbr Materials

PBR Brick wall: Ambient Occlusion (256px)

Ambient-occlusion map darkening crevices where ambient light is blocked. Part of the brick wall PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brick wall: Base Colour (Albedo) (256px)

Base-colour / albedo map: the diffuse reflectance with no lighting baked in. Part of the brick wall PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brick wall: Height / Displacement (256px)

Height / displacement map used for parallax or true displacement of the surface. Part of the brick wall PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brick wall: Metallic (256px)

Metallic map (black = dielectric, white = metal) selecting the reflectance model per texel. Part of the brick wall PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brick wall: Normal (256px)

Tangent-space normal map encoded from the surface height. Perturbs shading normals for detail without extra geometry. Part of the brick wall PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing+1· Conversion set

PBR Brick wall: Roughness (256px)

Roughness map (black = glossy, white = matte) driving the microfacet spread of the BRDF. Part of the brick wall PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brushed metal: Ambient Occlusion (256px)

Ambient-occlusion map darkening crevices where ambient light is blocked. Part of the brushed metal PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brushed metal: Base Colour (Albedo) (256px)

Base-colour / albedo map: the diffuse reflectance with no lighting baked in. Part of the brushed metal PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brushed metal: Height / Displacement (256px)

Height / displacement map used for parallax or true displacement of the surface. Part of the brushed metal PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brushed metal: Metallic (256px)

Metallic map (black = dielectric, white = metal) selecting the reflectance model per texel. Part of the brushed metal PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Brushed metal: Normal (256px)

Tangent-space normal map encoded from the surface height. Perturbs shading normals for detail without extra geometry. Part of the brushed metal PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing+1· Conversion set

PBR Brushed metal: Roughness (256px)

Roughness map (black = glossy, white = matte) driving the microfacet spread of the BRDF. Part of the brushed metal PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Ceramic tiles: Ambient Occlusion (256px)

Ambient-occlusion map darkening crevices where ambient light is blocked. Part of the ceramic tiles PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Ceramic tiles: Base Colour (Albedo) (256px)

Base-colour / albedo map: the diffuse reflectance with no lighting baked in. Part of the ceramic tiles PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Ceramic tiles: Height / Displacement (256px)

Height / displacement map used for parallax or true displacement of the surface. Part of the ceramic tiles PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Ceramic tiles: Metallic (256px)

Metallic map (black = dielectric, white = metal) selecting the reflectance model per texel. Part of the ceramic tiles PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Ceramic tiles: Normal (256px)

Tangent-space normal map encoded from the surface height. Perturbs shading normals for detail without extra geometry. Part of the ceramic tiles PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing+1· Conversion set

PBR Ceramic tiles: Roughness (256px)

Roughness map (black = glossy, white = matte) driving the microfacet spread of the BRDF. Part of the ceramic tiles PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Wood planks: Ambient Occlusion (256px)

Ambient-occlusion map darkening crevices where ambient light is blocked. Part of the wood planks PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Wood planks: Base Colour (Albedo) (256px)

Base-colour / albedo map: the diffuse reflectance with no lighting baked in. Part of the wood planks PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Wood planks: Height / Displacement (256px)

Height / displacement map used for parallax or true displacement of the surface. Part of the wood planks PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Wood planks: Metallic (256px)

Metallic map (black = dielectric, white = metal) selecting the reflectance model per texel. Part of the wood planks PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Wood planks: Normal (256px)

Tangent-space normal map encoded from the surface height. Perturbs shading normals for detail without extra geometry. Part of the wood planks PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing+1· Conversion set

PBR Wood planks: Roughness (256px)

Roughness map (black = glossy, white = matte) driving the microfacet spread of the BRDF. Part of the wood planks PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Woven fabric: Ambient Occlusion (256px)

Ambient-occlusion map darkening crevices where ambient light is blocked. Part of the woven fabric PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Woven fabric: Base Colour (Albedo) (256px)

Base-colour / albedo map: the diffuse reflectance with no lighting baked in. Part of the woven fabric PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Woven fabric: Height / Displacement (256px)

Height / displacement map used for parallax or true displacement of the surface. Part of the woven fabric PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Woven fabric: Metallic (256px)

Metallic map (black = dielectric, white = metal) selecting the reflectance model per texel. Part of the woven fabric PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

PBR Woven fabric: Normal (256px)

Tangent-space normal map encoded from the surface height. Perturbs shading normals for detail without extra geometry. Part of the woven fabric PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing+1· Conversion set

PBR Woven fabric: Roughness (256px)

Roughness map (black = glossy, white = matte) driving the microfacet spread of the BRDF. Part of the woven fabric PBR set. Pair with the other channels in this group.

File: PNG · Pbr Materials · 256 × 256 px
Use case: PBR material testingTexture map testing· Conversion set

Placeholder

Placeholder 1200×630 (og-card)

A 1200×630 placeholder image with its dimensions labelled on a coloured grid: a drop-in stand-in for testing layouts, responsive images, and thumbnail generation. Part of an avatar-to-Full-HD placeholder pack.

File: PNG · Placeholder · 1200 × 630 px
Use case: Resolution testingThumbnail generation· Conversion set

png

2.6 KB

Placeholder 150×150 (avatar)

A 150×150 placeholder image with its dimensions labelled on a coloured grid: a drop-in stand-in for testing layouts, responsive images, and thumbnail generation. Part of an avatar-to-Full-HD placeholder pack.

File: PNG · Placeholder · 150 × 150 px
Use case: Resolution testingThumbnail generation· Conversion set

Placeholder 1920×1080 (full-hd)

A 1920×1080 placeholder image with its dimensions labelled on a coloured grid: a drop-in stand-in for testing layouts, responsive images, and thumbnail generation. Part of an avatar-to-Full-HD placeholder pack.

File: PNG · Placeholder · 1920 × 1080 px
Use case: Resolution testingThumbnail generation· Conversion set

Placeholder 320×320 (square)

A 320×320 placeholder image with its dimensions labelled on a coloured grid: a drop-in stand-in for testing layouts, responsive images, and thumbnail generation. Part of an avatar-to-Full-HD placeholder pack.

File: PNG · Placeholder · 320 × 320 px
Use case: Resolution testingThumbnail generation· Conversion set

Placeholder 640×480 (vga)

A 640×480 placeholder image with its dimensions labelled on a coloured grid: a drop-in stand-in for testing layouts, responsive images, and thumbnail generation. Part of an avatar-to-Full-HD placeholder pack.

File: PNG · Placeholder · 640 × 480 px
Use case: Resolution testingThumbnail generation· Conversion set

Placeholder 800×600 (svga)

A 800×600 placeholder image with its dimensions labelled on a coloured grid: a drop-in stand-in for testing layouts, responsive images, and thumbnail generation. Part of an avatar-to-Full-HD placeholder pack.

File: PNG · Placeholder · 800 × 600 px
Use case: Resolution testingThumbnail generation· Conversion set

Raster Formats

A barista in her twenties (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate owner-cafe.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A Black man in his seventies (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-market-sen.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (avif lossy)

The same 512x350 photograph written as avif lossy. AV1 still image: modern, small, and unsupported by a surprising amount of tooling. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: AVIF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (bmp 24bit)

The same 512x350 photograph written as bmp 24bit. BMP stores rows bottom-up, so a reader that ignores that returns the image flipped. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: BMP · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (gif 256)

The same 512x350 photograph written as gif 256. GIF cannot hold more than 256 colours, so this is quantised by the format itself. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: GIF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (heic)

The same 512x350 photograph written as heic. What an iPhone produces by default, and what a lot of pipelines still cannot read. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: HEIC · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (ico multisize)

The same 512x350 photograph written as ico multisize. One file holding four resolutions; a reader that takes the first gets 16 pixels. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: ICO · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (jp2 lossless)

The same 512x350 photograph written as jp2 lossless. Wavelet coding, reversible mode; shares an extension family with the lossy variant below. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: JP2 · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (jp2 lossy)

The same 512x350 photograph written as jp2 lossy. Irreversible wavelet coding. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: JP2 · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (jpeg progressive)

The same 512x350 photograph written as jpeg progressive. Progressive scan order, which a naive reader renders as a blurry first pass or not at all. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: JPG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (jpeg q60)

The same 512x350 photograph written as jpeg q60. Baseline JPEG at quality 60. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: JPG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (jpeg q92)

The same 512x350 photograph written as jpeg q92. Baseline JPEG at quality 92. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: JPG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (pgm binary)

The same 512x350 photograph written as pgm binary. Greyscale Netpbm. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: PGM · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (png 24bit)

The same 512x350 photograph written as png 24bit. The lossless reference for this group; every lossy member is scored against it. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: PNG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (png grayscale)

The same 512x350 photograph written as png grayscale. Single channel, so a reader assuming three will read past the end of a row. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: PNG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (png interlaced)

The same 512x350 photograph written as png interlaced. Adam7 interlacing decodes in seven passes rather than one, which is a separate code path. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: PNG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (png palette256)

The same 512x350 photograph written as png palette256. A palette PNG carries indices and a colour table, not RGB pixels. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: PNG · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (ppm binary)

The same 512x350 photograph written as ppm binary. Netpbm binary: a tiny header then raw bytes, with no compression at all. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: PPM · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (tga rle)

The same 512x350 photograph written as tga rle. Run-length encoded Targa. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: TGA · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (tiff 16bit gray)

The same 512x350 photograph written as tiff 16bit gray. 16 bits per sample: a reader that truncates to 8 loses half the range silently. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: TIFF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (tiff deflate)

The same 512x350 photograph written as tiff deflate. Deflate inside TIFF. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: TIFF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (tiff jpeg)

The same 512x350 photograph written as tiff jpeg. A lossy JPEG stream inside a TIFF container, which breaks the assumption that TIFF is lossless. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: TIFF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (tiff lzw)

The same 512x350 photograph written as tiff lzw. LZW, the TIFF compression most readers assume. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: TIFF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (tiff uncompressed)

The same 512x350 photograph written as tiff uncompressed. Raw strips, the simplest TIFF and the largest. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: TIFF · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (webp lossless)

The same 512x350 photograph written as webp lossless. Lossless WebP, a different bitstream from the lossy one. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: WEBP · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black man in his seventies with grey (webp lossy)

The same 512x350 photograph written as webp lossy. Lossy WebP. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate p-market-sen.

File: WEBP · Raster Formats · 512 × 350 px
Use case: Media handling· Conversion set

A Black woman hairdresser in (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-salon-mid.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A cappuccino with latte art in (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate receipt-counter.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A cluttered joinery workshop (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-workshop.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A delivery driver with a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-driver-eve.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A delivery note on a clipboard (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate d-clipboard.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A hairdresser in her forties (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate owner-salon.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A market stall vendor (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate owner-market.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A Middle Eastern man chef (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-kitchen-chef.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A person's hands signing a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate office-desk.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A plated main course on a, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate plate-food.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

A plated main course on a, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-plate.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

A plus size white woman in her (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-shop-plus.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate d-invoice.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A printed invoice on a wooden, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate invoice-desk.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate d-menu.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A restaurant menu card standing, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate menu-table.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A scatter of crumpled paper (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate d-receipts.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A smiling woman with short (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate owner-restaurant.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A South Asian woman (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-trades-w.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

A warehouse worker holding a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate warehouse-clip.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

A wheelchair user working at (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-desk-chair.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

Abstract album cover artwork (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate c-album-abs.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, coral (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-ph-mesh2.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract flowing mesh gradient, violet (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-mesh-grad.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Abstract warm bokeh lights out (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-bokeh-warm.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Amateur football match on a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate sp-football.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

An East Asian man in his (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-ware-young.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

An electrician in work clothes (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate owner-trades.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

An empty modern open plan (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-office.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

An older man with a white (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate owner-shop.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

An older woman with a walking (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-stick-cafe.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate d-ledger.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

An open paper accounts ledger, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate ledger-book.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

APNG: Animated Spinner

A 12-frame animated PNG (APNG) of a rotating arc: a lossless, alpha-capable alternative to animated GIF. Useful for testing APNG support, frame extraction, and GIF↔APNG conversion.

File: APNG · Raster Formats · 128 × 128 px
Use case: Conversion testing

Arcade racing game track (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-racing-track.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Articulated lorry at a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-truck-artic.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Black woman mechanic in (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-mechanic-w.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Book cover artwork, empty rain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate c-book-thrill.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, misty (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate c-book-nature.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

Book cover artwork, weathered (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate c-book-hist.

File: WEBP · Raster Formats · 832 × 1216 px
Use case: Media handling· Conversion set

Brown horse standing in a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate n-horse-field.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Canvas backpack on a plain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-backpack.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

City bicycle leaning against a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-bike-city.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

City bus at a bus stop, grey (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-bus-city.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

City skyline across a river at (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-city-sky.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Climber on an indoor (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate sp-climbing.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up brown leather grain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-leather.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Close up indigo denim fabric (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-denim.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coarse recycled (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-paper2.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Close up of coffee being (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate f-coffee-pour.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Close up woven linen fabric (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-linen.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Colourful stylised platformer (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-platformer.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Commuter train at a station (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-train-plat.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Concrete spiral stairwell seen (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-stairwell.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark dungeon interior video (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-dungeon.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Dark moody blurred background, deep (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-ph-dark1.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Dark moody blurred background, near (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-ph-dark2.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Dark saloon car on a wet city (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-car-night.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

East african woman cleaner (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-cleaner-eve.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Electric scooter parked on a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-scooter.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty cafe interior before (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-cafe-empty.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty open plan office with (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-office-open.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty outdoor market stall (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-market-stall.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Empty stainless commercial (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-kitchen-clean.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-fantasy-keep.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (avif lossy)

The same 512x512 photograph written as avif lossy. AV1 still image: modern, small, and unsupported by a surprising amount of tooling. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: AVIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (bmp 24bit)

The same 512x512 photograph written as bmp 24bit. BMP stores rows bottom-up, so a reader that ignores that returns the image flipped. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: BMP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (gif 256)

The same 512x512 photograph written as gif 256. GIF cannot hold more than 256 colours, so this is quantised by the format itself. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: GIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (heic)

The same 512x512 photograph written as heic. What an iPhone produces by default, and what a lot of pipelines still cannot read. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: HEIC · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (ico multisize)

The same 512x512 photograph written as ico multisize. One file holding four resolutions; a reader that takes the first gets 16 pixels. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: ICO · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (jp2 lossless)

The same 512x512 photograph written as jp2 lossless. Wavelet coding, reversible mode; shares an extension family with the lossy variant below. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (jp2 lossy)

The same 512x512 photograph written as jp2 lossy. Irreversible wavelet coding. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (jpeg progressive)

The same 512x512 photograph written as jpeg progressive. Progressive scan order, which a naive reader renders as a blurry first pass or not at all. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (jpeg q60)

The same 512x512 photograph written as jpeg q60. Baseline JPEG at quality 60. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (jpeg q92)

The same 512x512 photograph written as jpeg q92. Baseline JPEG at quality 92. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (pgm binary)

The same 512x512 photograph written as pgm binary. Greyscale Netpbm. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: PGM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (png 24bit)

The same 512x512 photograph written as png 24bit. The lossless reference for this group; every lossy member is scored against it. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (png grayscale)

The same 512x512 photograph written as png grayscale. Single channel, so a reader assuming three will read past the end of a row. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (png interlaced)

The same 512x512 photograph written as png interlaced. Adam7 interlacing decodes in seven passes rather than one, which is a separate code path. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (png palette256)

The same 512x512 photograph written as png palette256. A palette PNG carries indices and a colour table, not RGB pixels. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (ppm binary)

The same 512x512 photograph written as ppm binary. Netpbm binary: a tiny header then raw bytes, with no compression at all. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: PPM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (tga rle)

The same 512x512 photograph written as tga rle. Run-length encoded Targa. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: TGA · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (tiff 16bit gray)

The same 512x512 photograph written as tiff 16bit gray. 16 bits per sample: a reader that truncates to 8 loses half the range silently. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (tiff deflate)

The same 512x512 photograph written as tiff deflate. Deflate inside TIFF. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (tiff jpeg)

The same 512x512 photograph written as tiff jpeg. A lossy JPEG stream inside a TIFF container, which breaks the assumption that TIFF is lossless. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (tiff lzw)

The same 512x512 photograph written as tiff lzw. LZW, the TIFF compression most readers assume. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (tiff uncompressed)

The same 512x512 photograph written as tiff uncompressed. Raw strips, the simplest TIFF and the largest. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (webp lossless)

The same 512x512 photograph written as webp lossless. Lossless WebP, a different bitstream from the lossy one. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Fantasy video game environment, stone (webp lossy)

The same 512x512 photograph written as webp lossy. Lossy WebP. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate g-fantasy-keep.

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Filipino woman nurse at a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-nurse-clinic.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Five colleagues of different (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-meeting-mix.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

Flat concrete wall texture (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-concrete.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat dark slate stone surface (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-slate.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat fine sand surface with (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-sand.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat lay of textured handmade (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-paper.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat oak wood grain surface (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-wood.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat red brick wall (avif lossy)

The same 512x341 photograph written as avif lossy. AV1 still image: modern, small, and unsupported by a surprising amount of tooling. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: AVIF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (bmp 24bit)

The same 512x341 photograph written as bmp 24bit. BMP stores rows bottom-up, so a reader that ignores that returns the image flipped. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: BMP · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (gif 256)

The same 512x341 photograph written as gif 256. GIF cannot hold more than 256 colours, so this is quantised by the format itself. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: GIF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (heic)

The same 512x341 photograph written as heic. What an iPhone produces by default, and what a lot of pipelines still cannot read. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: HEIC · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (ico multisize)

The same 512x341 photograph written as ico multisize. One file holding four resolutions; a reader that takes the first gets 16 pixels. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: ICO · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (jp2 lossless)

The same 512x341 photograph written as jp2 lossless. Wavelet coding, reversible mode; shares an extension family with the lossy variant below. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: JP2 · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (jp2 lossy)

The same 512x341 photograph written as jp2 lossy. Irreversible wavelet coding. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: JP2 · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (jpeg progressive)

The same 512x341 photograph written as jpeg progressive. Progressive scan order, which a naive reader renders as a blurry first pass or not at all. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: JPG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (jpeg q60)

The same 512x341 photograph written as jpeg q60. Baseline JPEG at quality 60. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: JPG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (jpeg q92)

The same 512x341 photograph written as jpeg q92. Baseline JPEG at quality 92. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: JPG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (pgm binary)

The same 512x341 photograph written as pgm binary. Greyscale Netpbm. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: PGM · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (png 24bit)

The same 512x341 photograph written as png 24bit. The lossless reference for this group; every lossy member is scored against it. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (png grayscale)

The same 512x341 photograph written as png grayscale. Single channel, so a reader assuming three will read past the end of a row. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (png interlaced)

The same 512x341 photograph written as png interlaced. Adam7 interlacing decodes in seven passes rather than one, which is a separate code path. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (png palette256)

The same 512x341 photograph written as png palette256. A palette PNG carries indices and a colour table, not RGB pixels. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (ppm binary)

The same 512x341 photograph written as ppm binary. Netpbm binary: a tiny header then raw bytes, with no compression at all. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: PPM · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (tga rle)

The same 512x341 photograph written as tga rle. Run-length encoded Targa. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: TGA · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (tiff 16bit gray)

The same 512x341 photograph written as tiff 16bit gray. 16 bits per sample: a reader that truncates to 8 loses half the range silently. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (tiff deflate)

The same 512x341 photograph written as tiff deflate. Deflate inside TIFF. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (tiff jpeg)

The same 512x341 photograph written as tiff jpeg. A lossy JPEG stream inside a TIFF container, which breaks the assumption that TIFF is lossless. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (tiff lzw)

The same 512x341 photograph written as tiff lzw. LZW, the TIFF compression most readers assume. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (tiff uncompressed)

The same 512x341 photograph written as tiff uncompressed. Raw strips, the simplest TIFF and the largest. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (webp lossless)

The same 512x341 photograph written as webp lossless. Lossless WebP, a different bitstream from the lossy one. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: WEBP · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (webp lossy)

The same 512x341 photograph written as webp lossy. Lossy WebP. Every member of this group is identical pixels, so a decoder returning different dimensions or a different pixel count has lost something rather than simply produced a different file; the 24-bit PNG in this group is lossless and is the reference the lossy members are measured against Source plate t-tex-brick.

File: WEBP · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Flat red brick wall (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-brick.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat rusted steel plate (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-rust.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flat white marble slab with (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-tex-marble.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Flock of sheep on a green (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate n-sheep-hill.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Four colleagues around a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate meeting-table.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

Glass water bottle on a plain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-bottle.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Golden retriever sitting on (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate n-dog-park.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Green farm tractor in a field (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-tractor.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Group of road cyclists on a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate sp-cyclists.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a bearded (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-m3.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (avif lossy)

File: AVIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (bmp 24bit)

File: BMP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (gif 256)

File: GIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (heic)

File: HEIC · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (ico multisize)

File: ICO · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (jp2 lossless)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (jp2 lossy)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (jpeg progressive)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (jpeg q60)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (jpeg q92)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (pgm binary)

File: PGM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (png 24bit)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (png grayscale)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (png interlaced)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (png palette256)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (ppm binary)

File: PPM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (tga rle)

File: TGA · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (tiff 16bit gray)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (tiff deflate)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (tiff jpeg)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (tiff lzw)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (tiff uncompressed)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (webp lossless)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (webp lossy)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a black (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-f3.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a middle (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-m6.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a south (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-m4.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-m5.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of a, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-nb1.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

webp

99.1 KB

Head and shoulders portrait of a, plate C (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-avatar-f1.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

webp

62.8 KB

Head and shoulders portrait of a, plate D (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-avatar-f2.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

webp

64.8 KB

Head and shoulders portrait of a, plate E (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-avatar-m1.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an east (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-f5.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Head and shoulders portrait of an, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-av-f4.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

webp

87.7 KB

Head and shoulders portrait of an, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate a-avatar-m2.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

ICO: Multi-Resolution App Icon

A Windows ICO containing four square sizes (16/32/48/64 px) of a simple app glyph: the classic favicon/desktop-icon container. Handy for testing icon extraction and multi-size rendering.

File: ICO · Raster Formats
Use case: Conversion testing

Isometric stylised village (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-isometric.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Latina teacher beside a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-teacher-cls.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Low poly stylised island with (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-lowpoly-isle.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged indian man (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-accountant.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle aged white male butcher (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-butcher-mid.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Middle eastern man tailor (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-tailor-mid.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Modern minimalist living room (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-interior-mod.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Moody album cover artwork (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate c-album-moody.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Mountain valley with a river (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-mountain.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Mud splattered pickup truck on (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-pickup-mud.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Narrow european old town (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-street-eu.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Older white man farmer (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-farmer-field.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Open laptop on a plain white (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-laptop.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Over ear headphones on a plain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-headphone.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead colourful garden (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate f-plate-salad.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate f-plate-pasta.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (avif lossy)

File: AVIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (bmp 24bit)

File: BMP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (gif 256)

File: GIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (heic)

File: HEIC · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (ico multisize)

File: ICO · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (jp2 lossless)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (jp2 lossy)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (jpeg progressive)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (jpeg q60)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (jpeg q92)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (pgm binary)

File: PGM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (png 24bit)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (png grayscale)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (png interlaced)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (png palette256)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (ppm binary)

File: PPM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (tga rle)

File: TGA · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (tiff 16bit gray)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (tiff deflate)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (tiff jpeg)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (tiff lzw)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (tiff uncompressed)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (webp lossless)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Overhead plate of fresh pasta with (webp lossy)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Passenger aircraft at an (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-plane-gate.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Path through a deciduous (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-forest-path.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

PBM: Netpbm Test Image (P4)

A 256×256 fruit still-life image as binary Netpbm P4 (1-bit bitmap): the minimal, header-plus-raw-pixels family used across Unix imaging tools. For testing Netpbm parsers and conversion.

File: PBM · Raster Formats · 256 × 256 px
Use case: Conversion testing

Person lifting weights in a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate sp-gym.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

PGM: Netpbm Test Image (P5)

A 256×256 fruit still-life image as binary Netpbm P5 (8-bit greyscale): the minimal, header-plus-raw-pixels family used across Unix imaging tools. For testing Netpbm parsers and conversion.

File: PGM · Raster Formats · 256 × 256 px
Use case: Conversion testing

Pixel art side scrolling game (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-pixel-scene.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Potted houseplant on a plain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-plant-pot.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

PPM: Netpbm Test Image (P6)

A 256×256 fruit still-life image as binary Netpbm P6 (24-bit colour): the minimal, header-plus-raw-pixels family used across Unix imaging tools. For testing Netpbm parsers and conversion.

File: PPM · Raster Formats · 256 × 256 px
Use case: Conversion testing

Red vintage convertible parked (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-car-classic.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Retro album cover artwork (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate c-album-retro.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Rocky coastline with breaking (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-coast.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Runner on a park path in the (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate sp-runner.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Rustic sourdough loaves on a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate f-bread-rustic.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Science fiction video game (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-scifi-corr.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Senior south asian woman (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-florist-sen.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-car-sedan.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (avif lossy)

File: AVIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (bmp 24bit)

File: BMP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (gif 256)

File: GIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (heic)

File: HEIC · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (ico multisize)

File: ICO · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (jp2 lossless)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (jp2 lossy)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (jpeg progressive)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (jpeg q60)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (jpeg q92)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (pgm binary)

File: PGM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (png 24bit)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (png grayscale)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (png interlaced)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (png palette256)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (ppm binary)

File: PPM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (tga rle)

File: TGA · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (tiff 16bit gray)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (tiff deflate)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (tiff jpeg)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (tiff lzw)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (tiff uncompressed)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (webp lossless)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Silver sedan parked on a suburban (webp lossy)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-shoe.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (avif lossy)

File: AVIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (bmp 24bit)

File: BMP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (gif 256)

File: GIF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (heic)

File: HEIC · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (ico multisize)

File: ICO · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (jp2 lossless)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (jp2 lossy)

File: JP2 · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (jpeg progressive)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (jpeg q60)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (jpeg q92)

File: JPG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (pgm binary)

File: PGM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (png 24bit)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (png grayscale)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (png interlaced)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (png palette256)

File: PNG · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (ppm binary)

File: PPM · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (tga rle)

File: TGA · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (tiff 16bit gray)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (tiff deflate)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (tiff jpeg)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (tiff lzw)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (tiff uncompressed)

File: TIFF · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (webp lossless)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Single running shoe on a plain white (webp lossy)

File: WEBP · Raster Formats · 512 × 512 px
Use case: Media handling· Conversion set

Small fishing boat moored in a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-boat-harbour.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Small garden birds on a wooden (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate n-birds-feed.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Small independent grocery (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-shop-front.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Smooth abstract gradient (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-grad-mono.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-grad-cool.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (avif lossy)

File: AVIF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (bmp 24bit)

File: BMP · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (gif 256)

File: GIF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (heic)

File: HEIC · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (ico multisize)

File: ICO · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (jp2 lossless)

File: JP2 · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (jp2 lossy)

File: JP2 · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (jpeg progressive)

File: JPG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (jpeg q60)

File: JPG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (jpeg q92)

File: JPG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (pgm binary)

File: PGM · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (png 24bit)

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (png grayscale)

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (png interlaced)

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (png palette256)

File: PNG · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (ppm binary)

File: PPM · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (tga rle)

File: TGA · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (tiff 16bit gray)

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (tiff deflate)

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (tiff jpeg)

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (tiff lzw)

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (tiff uncompressed)

File: TIFF · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (webp lossless)

File: WEBP · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, deep teal to (webp lossy)

File: WEBP · Raster Formats · 512 × 341 px
Use case: Media handling· Conversion set

Smooth abstract gradient, warm (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-grad-warm.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Smooth vertical gradient from (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-grad-mono2.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-ph-soft3.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, pale (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-ph-soft1.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Soft blurred abstract background, warm (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate t-ph-soft2.

File: WEBP · Raster Formats · 1280 × 853 px
Use case: Media handling· Conversion set

Sports motorcycle parked at a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-moto-sport.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Steel truss bridge over a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate l-bridge.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy knight (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-char-knight.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised fantasy mage (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-char-mage.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Stylised science fiction (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-char-scifi.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Swimmer mid stroke in an (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate sp-swimmer.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Tabby cat sitting on a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate n-cat-window.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

TGA: Truevision Test Image

A 256×256 fruit still-life image saved as uncompressed Truevision TGA: a format common in games and 3D texturing. For testing TGA decoders and conversion to modern formats.

File: TGA · Raster Formats · 256 × 256 px
Use case: Conversion testing

The exterior of a small, plate A (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-bakery.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

The exterior of a small, plate B (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate storefront.

File: WEBP · Raster Formats · 1216 × 832 px
Use case: Media handling· Conversion set

Tropical fish in a planted (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate n-fish-tank.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Voxel art city block, blocky (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate g-voxel-city.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Warehouse aisle between tall (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate s-warehouse-a.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Wheelchair user chef at a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-chef-wheel.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

White electric hatchback at a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-car-ev.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

White panel delivery van (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-van-delivery.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Wristwatch on a plain grey (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate pr-watch.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Yellow forklift in a warehouse (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate v-forklift.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Young east asian woman baker (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-baker-young.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Young man barista signing to a (WebP q85)

The same plate as WebP at quality 85, for format-handling coverage beside the PNG and JPEG encodings of identical pixels Plate p-barista-deaf.

File: WEBP · Raster Formats · 1024 × 1024 px
Use case: Media handling· Conversion set

Resolution Ladder

Resolution: 1080×1920 (vertical 9:16)

A fruit still life rendered at 1080×1920 (vertical 9:16) with a faint registration grid: part of a ladder from 16px to 4K for testing scaling, thumbnail generation, and responsive layout on recognisable content.

File: PNG · Resolution Ladder · 1080 × 1920 px
Use case: Resolution testingThumbnail generation· Conversion set

Resolution: 16×16 (16px thumbnail)

A fruit still life rendered at 16×16 (16px thumbnail) with a faint registration grid: part of a ladder from 16px to 4K for testing scaling, thumbnail generation, and responsive layout on recognisable content.

File: PNG · Resolution Ladder · 16 × 16 px
Use case: Resolution testingThumbnail generation· Conversion set

Resolution: 1920×1080 (1080p 16:9)

A fruit still life rendered at 1920×1080 (1080p 16:9) with a faint registration grid: part of a ladder from 16px to 4K for testing scaling, thumbnail generation, and responsive layout on recognisable content.

File: PNG · Resolution Ladder · 1920 × 1080 px
Use case: Resolution testingThumbnail generation· Conversion set

Resolution: 256×256 (256px)

A fruit still life rendered at 256×256 (256px) with a faint registration grid: part of a ladder from 16px to 4K for testing scaling, thumbnail generation, and responsive layout on recognisable content.

File: PNG · Resolution Ladder · 256 × 256 px
Use case: Resolution testingThumbnail generation· Conversion set

Resolution: 2560×1440 (1440p 16:9)

A fruit still life rendered at 2560×1440 (1440p 16:9) with a faint registration grid: part of a ladder from 16px to 4K for testing scaling, thumbnail generation, and responsive layout on recognisable content.

File: PNG · Resolution Ladder · 2560 × 1440 px
Use case: Resolution testingThumbnail generation· Conversion set

Resolution: 3840×2160 (4K 16:9)

A fruit still life rendered at 3840×2160 (4K 16:9) with a faint registration grid: part of a ladder from 16px to 4K for testing scaling, thumbnail generation, and responsive layout on recognisable content.

File: PNG · Resolution Ladder · 3840 × 2160 px
Use case: Resolution testingThumbnail generation· Conversion set

Restoration Set

Restoration Clean: Crease (512px)

Clean reference for the crease photo-restoration case. Diff a restorer's output against this file; the damaged input and mask share group img-restore-crease.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Clean: Dust (512px)

Clean reference for the dust photo-restoration case. Diff a restorer's output against this file; the damaged input and mask share group img-restore-dust.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Clean: Horizontal Cut (512px)

Clean reference for the horizontal-cut photo-restoration case. Diff a restorer's output against this file; the damaged input and mask share group img-restore-horizontal-cut.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Clean: Scratch (512px)

Clean reference for the scratch photo-restoration case. Diff a restorer's output against this file; the damaged input and mask share group img-restore-scratch.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Clean: Torn Corner (512px)

Clean reference for the torn-corner photo-restoration case. Diff a restorer's output against this file; the damaged input and mask share group img-restore-torn-corner.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Clean: Water Stain (512px)

Clean reference for the water-stain photo-restoration case. Diff a restorer's output against this file; the damaged input and mask share group img-restore-water-stain.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Damaged: Crease (512px)

Damaged input for crease restoration testing (seed 408). Pair with the clean reference and damage mask in this group.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Damaged: Dust (512px)

Damaged input for dust restoration testing (seed 410). Pair with the clean reference and damage mask in this group.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Damaged: Horizontal Cut (512px)

Damaged input for horizontal-cut restoration testing (seed 413). Pair with the clean reference and damage mask in this group.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Damaged: Scratch (512px)

Damaged input for scratch restoration testing (seed 409). Pair with the clean reference and damage mask in this group.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Damaged: Torn Corner (512px)

Damaged input for torn-corner restoration testing (seed 412). Pair with the clean reference and damage mask in this group.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Damaged: Water Stain (512px)

Damaged input for water-stain restoration testing (seed 411). Pair with the clean reference and damage mask in this group.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationImage pipeline QA· Conversion set

Restoration Mask: Crease (512px)

Black-and-white damage mask for the crease restoration case (white = damaged). Use with the damaged input and clean reference.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationInpainting· Conversion set

Restoration Mask: Dust (512px)

Black-and-white damage mask for the dust restoration case (white = damaged). Use with the damaged input and clean reference.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationInpainting· Conversion set

Restoration Mask: Horizontal Cut (512px)

Black-and-white damage mask for the horizontal-cut restoration case (white = damaged). Use with the damaged input and clean reference.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationInpainting· Conversion set

Restoration Mask: Scratch (512px)

Black-and-white damage mask for the scratch restoration case (white = damaged). Use with the damaged input and clean reference.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationInpainting· Conversion set

Restoration Mask: Torn Corner (512px)

Black-and-white damage mask for the torn-corner restoration case (white = damaged). Use with the damaged input and clean reference.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationInpainting· Conversion set

Restoration Mask: Water Stain (512px)

Black-and-white damage mask for the water-stain restoration case (white = damaged). Use with the damaged input and clean reference.

File: PNG · Restoration Set · 512 × 512 px
Use case: Photo restorationInpainting· Conversion set

Seg Set

Segmentation Scene 01: Class Map (JSON)

Class-id → label map for segmentation scene 1. Ground truth companion to the RGB and mask.

File: JSON · Seg Set
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 01: Class Mask (288px)

Indexed L mask for scene 1: pixel value equals class id (1=sky, 2=ground, 3=object, 4=accent).

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 01: RGB (288px)

Synthetic multi-class scene 1 (sky/ground/object/accent) for semantic-segmentation smoke tests. Paired with an indexed mask and class JSON.

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 02: Class Map (JSON)

Class-id → label map for segmentation scene 2. Ground truth companion to the RGB and mask.

File: JSON · Seg Set
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 02: Class Mask (288px)

Indexed L mask for scene 2: pixel value equals class id (1=sky, 2=ground, 3=object, 4=accent).

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 02: RGB (288px)

Synthetic multi-class scene 2 (sky/ground/object/accent) for semantic-segmentation smoke tests. Paired with an indexed mask and class JSON.

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 03: Class Map (JSON)

Class-id → label map for segmentation scene 3. Ground truth companion to the RGB and mask.

File: JSON · Seg Set
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 03: Class Mask (288px)

Indexed L mask for scene 3: pixel value equals class id (1=sky, 2=ground, 3=object, 4=accent).

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 03: RGB (288px)

Synthetic multi-class scene 3 (sky/ground/object/accent) for semantic-segmentation smoke tests. Paired with an indexed mask and class JSON.

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 04: Class Map (JSON)

Class-id → label map for segmentation scene 4. Ground truth companion to the RGB and mask.

File: JSON · Seg Set
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 04: Class Mask (288px)

Indexed L mask for scene 4: pixel value equals class id (1=sky, 2=ground, 3=object, 4=accent).

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 04: RGB (288px)

Synthetic multi-class scene 4 (sky/ground/object/accent) for semantic-segmentation smoke tests. Paired with an indexed mask and class JSON.

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 05: Class Map (JSON)

Class-id → label map for segmentation scene 5. Ground truth companion to the RGB and mask.

File: JSON · Seg Set
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 05: Class Mask (288px)

Indexed L mask for scene 5: pixel value equals class id (1=sky, 2=ground, 3=object, 4=accent).

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 05: RGB (288px)

Synthetic multi-class scene 5 (sky/ground/object/accent) for semantic-segmentation smoke tests. Paired with an indexed mask and class JSON.

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 06: Class Map (JSON)

Class-id → label map for segmentation scene 6. Ground truth companion to the RGB and mask.

File: JSON · Seg Set
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 06: Class Mask (288px)

Indexed L mask for scene 6: pixel value equals class id (1=sky, 2=ground, 3=object, 4=accent).

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Segmentation Scene 06: RGB (288px)

Synthetic multi-class scene 6 (sky/ground/object/accent) for semantic-segmentation smoke tests. Paired with an indexed mask and class JSON.

File: PNG · Seg Set · 288 × 288 px
Use case: SegmentationComputer vision· Conversion set

Style Transfer

Style Transfer Photo Input (512px)

Fruit still-life content photo for neural / filter style-transfer tests. Pair with the style reference images in this group.

File: PNG · Style Transfer · 512 × 512 px
Use case: Style transferImage pipeline QA· Conversion set

Style Transfer Reference: Comic (512px)

Style reference (comic) derived from the same scene. Use as a target style when transferring the photo input.

File: PNG · Style Transfer · 512 × 512 px
Use case: Style transferImage pipeline QA· Conversion set

Style Transfer Reference: Geometric (512px)

Abstract geometric style plate for style-transfer tools that accept a separate style image.

File: PNG · Style Transfer · 512 × 512 px
Use case: Style transferImage pipeline QA· Conversion set

Style Transfer Reference: Line Art (512px)

Style reference (line-art) derived from the same scene. Use as a target style when transferring the photo input.

File: PNG · Style Transfer · 512 × 512 px
Use case: Style transferImage pipeline QA· Conversion set

Style Transfer Reference: Paint Like (512px)

Style reference (paint-like) derived from the same scene. Use as a target style when transferring the photo input.

File: PNG · Style Transfer · 512 × 512 px
Use case: Style transferImage pipeline QA· Conversion set

Style Transfer Reference: Posterize (512px)

Style reference (posterize) derived from the same scene. Use as a target style when transferring the photo input.

File: PNG · Style Transfer · 512 × 512 px
Use case: Style transferImage pipeline QA· Conversion set

Texture

Texture: Checkerboard

A 256×256 checkerboard texture that tiles seamlessly, for testing texture mapping, pattern fills, tiling, and background rendering.

File: PNG · Texture
Use case: Image pipeline QA· Conversion set

Texture: Diagonal Stripes

A 256×256 diagonal stripes texture that tiles seamlessly, for testing texture mapping, pattern fills, tiling, and background rendering.

File: PNG · Texture
Use case: Image pipeline QA· Conversion set

Texture: Grain

A 256×256 grain texture, for testing texture mapping, pattern fills, tiling, and background rendering.

File: PNG · Texture
Use case: Image pipeline QA· Conversion set

Texture: Grid

A 256×256 grid texture that tiles seamlessly, for testing texture mapping, pattern fills, tiling, and background rendering.

File: PNG · Texture
Use case: Image pipeline QA· Conversion set

Texture: Polka Dots

A 256×256 polka dots texture that tiles seamlessly, for testing texture mapping, pattern fills, tiling, and background rendering.

File: PNG · Texture
Use case: Image pipeline QA· Conversion set

Upscale Set

Upscale Input A: 128px Bicubic

Bicubic downscale to 128×128 from the 1024px reference. Use as a super-resolution input.

File: PNG · Upscale Set · 128 × 128 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Input A: 256px + JPEG q35

256×256 bicubic downscale further JPEG-compressed at quality 35: a harder super-resolution input with artefacts.

File: PNG · Upscale Set · 256 × 256 px
Use case: Super-resolutionCompression testing· Conversion set

Upscale Input A: 256px Bicubic

Bicubic downscale to 256×256 from the 1024px reference. Use as a super-resolution input.

File: PNG · Upscale Set · 256 × 256 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Input A: 64px Bicubic

Bicubic downscale to 64×64 from the 1024px reference. Use as a super-resolution input.

File: PNG · Upscale Set · 64 × 64 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Input B: 128px Bicubic

Bicubic downscale to 128×128 from the 1024px reference. Use as a super-resolution input.

File: PNG · Upscale Set · 128 × 128 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Input B: 256px + JPEG q35

256×256 bicubic downscale further JPEG-compressed at quality 35: a harder super-resolution input with artefacts.

File: PNG · Upscale Set · 256 × 256 px
Use case: Super-resolutionCompression testing· Conversion set

Upscale Input B: 256px Bicubic

Bicubic downscale to 256×256 from the 1024px reference. Use as a super-resolution input.

File: PNG · Upscale Set · 256 × 256 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Input B: 64px Bicubic

Bicubic downscale to 64×64 from the 1024px reference. Use as a super-resolution input.

File: PNG · Upscale Set · 64 × 64 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Reference A: 1024px

Clean 1024×1024 reference for super-resolution base A. Upscale the smaller inputs in this group toward this ground truth.

File: PNG · Upscale Set · 1024 × 1024 px
Use case: Super-resolutionResolution testing· Conversion set

Upscale Reference B: 1024px

Clean 1024×1024 reference for super-resolution base B. Upscale the smaller inputs in this group toward this ground truth.

File: PNG · Upscale Set · 1024 × 1024 px
Use case: Super-resolutionResolution testing· Conversion set

Vector

svg

547 B

Animated Spinner (SVG)

An animated loading spinner using SMIL (animateTransform): a self-contained animated SVG for testing whether a renderer or converter handles SVG animation, and how it rasterises an animated frame.

File: SVG · Vector
Use case: Conversion testingEditor testing· Conversion set

Bar Chart (SVG)

A bar chart drawn as vector SVG (axis, gridlines, labelled bars), the scalable counterpart to the raster chart images, for testing SVG chart rendering and vector conversion.

File: SVG · Vector
Use case: Conversion testingEditor testing+1· Conversion set

Flat Illustration (SVG)

A flat vector landscape (gradient sky, sun, layered hills, and trees built from bezier paths), a richer SVG for testing gradient and path rendering, thumbnailing, and vector conversion.

File: SVG · Vector
Use case: Conversion testingEditor testing+1· Conversion set

svg

751 B

Logo (SVG)

A generic sample brand logo as scalable vector graphics (a gradient mark plus a wordmark) for testing SVG rendering, rasterisation, and vector-to-raster conversion. Opens in the in-browser editor.

File: SVG · Vector
Use case: Conversion testingEditor testing+1· Conversion set

svg

532 B

UI Icon (SVG)

A single stroke-style UI icon (a document) as a 24×24 SVG: the kind of vector icon used in interfaces, for testing icon rendering, recolouring, and SVG-to-PNG conversion.

File: SVG · Vector
Use case: Conversion testingEditor testing· Conversion set

Video Thumbnails

Coast drift: scrub-preview sprite sheet

The scrub-preview sheet for the film in this group: 24 tiles in a 6x4 grid, one sampled per second, each 160x90. This is the form a player requests when a viewer drags the scrubber. A two-second clip yields two tiles and proves nothing about tiling; 24 seconds fills the grid.

File: JPG · Video Thumbnails
Use case: Media handling· Paired fixture