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.
Specifications
- Width
- 512
- Height
- 512
- Role
- result
- Removal Kind
- blob
- Mask Coverage
- 0.05614
- Source Plate
- nss-v-boat-harbour_00001_.png
- Crop Box
- 279,288,791,800
- Group Members
- 4
- Synthetic
- true
- Disclosure
- Derived from an AI-generated plate. Synthetic, not a photograph of a real scene.
- Schema Version
- 1
- Model
- Stable Diffusion 1.5 (v1-5-pruned-emaonly-fp16)
- Sampler
- dpmpp_2m / karras, 25 steps, CFG 7.0, denoise 1.0
- Seed
- 3018596753
- Grow Mask By
- 12
- Ring Px
- 16
- Inside Changed Mad
- 30.711
- Ring Changed Mad
- 9.217
- Outside Changed Mad
- 7.026
- Inside Vs Truth Mad
- 25.736
- Outside Vs Truth Mad
- 7.026
- Palette Gap
- 12.403
- Truth Palette Gap
- 8.198
- Alt Text
- Small fishing boat moored in a harbour, morning light
Testing contract
Expected to pass- Scenario
- Compare this file against the source and the clean plate in its group, measuring inside the mask, in the 16-pixel ring around it, and beyond that ring separately.
- Expected result
- Inside the mask it differs from the source by 30.711/255 - the object is gone. Beyond the ring it differs by only 7.026/255, the full-frame VAE round trip, so this was a local edit and not a repaint. Against the ground-truth plate the masked region differs by 25.736/255, and that number is NOT expected to be small: the masked latents are erased before sampling, so the model invented plausible content rather than recovering what was there. Measure the three zones separately or the feathered grow beyond the mask scores as damage.
What is a .png file?
PNG (Portable Network Graphics) is a raster image format using lossless DEFLATE compression. It supports full 8- or 16-bit-per-channel truecolor, palette, and greyscale modes with an optional alpha channel, but no animation. It is the standard choice for screenshots, logos, and graphics with sharp edges or transparency.
How to use this file
Use an example PNG to test image decoders, alpha-compositing, thumbnail generators, and format converters, or to verify that a pipeline preserves transparency and color depth on round-trip.
How to use this file for testing
“V boat harbour: object removed” is a deterministic Novus Examples fixture for Inpainting. Images with cut-out regions (rects, circles, corners, strips, irregular tears) and matching masks for testing inpainting, generative fill, content-aware fill, and object-removal pipelines.
Documented properties for this file: 512×512 · seed 3018596753 · result. Compare results against paired or grouped companions on this page when present (clean↔damaged, searchable↔scanned, or format twins) so scores stay reproducible across runs.
Download the file once, keep the path stable in CI or local scripts, and treat the spec table as the contract: dimensions, seeds, field lists, and roles are intentional. Corrupt or invalid samples are labelled as such, expect parsers to fail loudly rather than silently accept them.
For image AI or enhancement tools, run the model or filter on this file and diff against the clean or ground-truth companion in the same group when available. Keep seeds and documented damage parameters in your evaluation notes so regressions are attributable.
Run your enhancement model on the degraded input and diff against the clean ground-truth companion in the same group; keep the documented degradation parameters and seed in your eval notes so scores are reproducible.
Code examples
<img src="v-boat-harbour-cleaned.png" alt="Example image" width="640" loading="lazy">Related files
- pngA 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.
- pngA av f3: object composited inThe 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.
- pngA av f3: object removedHead 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.
- pngA av f3: removal maskThe 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.
- pngA 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.
- pngA av nb1: object composited inThe 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.
Generated by generation/comfy_removal_wave.py. Free for any use, no attribution required, license.