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.
Page 85 of 98; 24 results per page.
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.
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.
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.
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.
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.
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.
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.
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.
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.
Compact sparkline PNG for chart-extraction / thumbnail vision tests.
Compact sparkline PNG for chart-extraction / thumbnail vision tests.
Compact sparkline PNG for chart-extraction / thumbnail vision tests.
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 v-moto-sport.
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 v-moto-sport.
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 v-moto-sport.
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 v-moto-sport.
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.
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.
RGBA sprite sheet (2×2 cells of 64px) with numbered frames for atlas / CSS-sprite tools.
RGBA sprite sheet (3×2 cells of 64px) with numbered frames for atlas / CSS-sprite tools.
RGBA sprite sheet (4×1 cells of 64px) with numbered frames for atlas / CSS-sprite tools.
PNG whose colour intent is documented as sRGB in specs (no mandatory ICC blob), for tools that read metadata/specs.
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 l-bridge.
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 l-bridge.
