This dataset contains approximately 75k fake images, manipulated by text-guided inpainting methods (SD2, SDXL, and Adobe Firefly). The authentic images originate from MS-COCO, with a CC BY 4.0 license, and have resolutions up to 1024x1024 px. We provide both the manipulated image where the inpainted area is spliced in the original image (SD2-sp, PS-sp), as well as the fully-regenerated image (SD2-fr, SDXL-fr), when possible.
The dataset corresponds to the paper "TGIF: Text-Guided Inpainting Forgery Dataset", which is currently submitted and under review.
We distribute this dataset under the CC BY-SA 4.0 license.
- Separate download links for training/validation and test set
- Add GIQA scores to metadata
- Add benchmark results (per image)
- Add code used for generation (SD python script & Photoshop script)
- Add paper pdf and BibTex code
| Manipulation types | |
|---|---|
| # masks | 2 (segmentation & bounding box) |
| # variations | 3 per generation |
| # sub-datasets | 4 (SD2-sp, PS-sp, SD2-fr, SDXL-fr) |
| Total # manipulated images per authentic image | 2 * 3 * 4 = 24 |
| Dataset size | Training | Validation | Testing | Total |
|---|---|---|---|---|
| # authentic images | 2 440 | 341 | 343 | 3 124 |
| # manipulated images | 58 560 | 8 184 | 8 232 | 74 976 |
- Fake images
- SD2 (46 GB)
- SDXL (41 GB)
- Adobe Firefly (17.8 GB)
- Real images
- SD2 (4 GB)
- SDXL crops (3 GB)
- Masks
- Metadata