Find files, editable templates and browser test targets by what you need to make or test. The directory below is cut by format; the two collections under it cut the same library by subject and by workflow.
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.
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.
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.
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.
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.
Where the original view of a warehouse worker holding a clipboard beside tall racking, high bay lighting, boxes on pallets landed inside the expanded frame, as JSON: the rectangle 0,0 to 512,512 in a 704x512 canvas, the padding that produced it, and the 40-pixel feather that means the comparison must be inset. Authored BEFORE the expansion ran, so it is ground truth rather than a rectangle recovered afterwards by matching the original against the result - which is what a consumer would otherwise have to do, and which cannot distinguish a correct offset from a plausible one.
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.
A 2.0 second 640x480 H.264 clip at 24 fps, 49 frames. Animated from the still nss-warehouse-clip_00001_.png, so the first frame is a known image and frame extraction can be checked against it. Measured mean inter-frame change 0.0452 (active). Substantial frame-to-frame change, the end of the range where naive scene detection starts firing. Synthetic footage: two seconds at this size is a decoder and pipeline fixture rather than showcase material, hands degrade in later frames, and no text in shot is legible.
A warranty registration covering the fields that make post-purchase records awkward: a serial number with a checksum-shaped pattern, a purchase date bounded to the past, and a retailer list with an other option.
