Datasets:
Formats:
parquet
Languages:
English
Size:
100K - 1M
ArXiv:
Tags:
image-editing
instruction-based-editing
forgery-localization
manipulation-localization
benchmark
License:
Remove Known issues and robustness notes
Browse files
README.md
CHANGED
|
@@ -254,16 +254,6 @@ The paper defines two protocols and reports pixel-level and region-level metrics
|
|
| 254 |
- **Metrics.** Accuracy, AUC and F1 treat each pixel as a binary decision; Dice and
|
| 255 |
mIoU measure region overlap. Pixel-level accuracy/AUC are dominated by the large
|
| 256 |
untouched background, so Dice/mIoU are the more informative scores.
|
| 257 |
-
- **Robustness.** Evaluation can additionally be performed under JPEG compression
|
| 258 |
-
(quality 0.75) and Gaussian blurring (sigma = 1.0).
|
| 259 |
-
|
| 260 |
-
## Known issues
|
| 261 |
-
|
| 262 |
-
- For 12 samples in Qwen-Image-Edit, the source metadata omitted `img_name` and
|
| 263 |
-
`edit_type`. These fields were recovered from the filename (all 12 are
|
| 264 |
-
`addition` samples); no other field is affected.
|
| 265 |
-
- `boxes` / `scores` are detector/segmenter outputs and can include coordinates
|
| 266 |
-
slightly outside the image, or several boxes for one instruction.
|
| 267 |
|
| 268 |
## License and citation
|
| 269 |
|
|
|
|
| 254 |
- **Metrics.** Accuracy, AUC and F1 treat each pixel as a binary decision; Dice and
|
| 255 |
mIoU measure region overlap. Pixel-level accuracy/AUC are dominated by the large
|
| 256 |
untouched background, so Dice/mIoU are the more informative scores.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 257 |
|
| 258 |
## License and citation
|
| 259 |
|