| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Image Generation | CelebA 64 x 64 (test) | FID1.03 | 203 | |
| Facial Attribute Classification | CelebA | Accuracy93.22 | 163 | |
| Classification | CelebA | Avg Accuracy94.9 | 137 | |
| Image Generation | CelebA | FID2.71 | 110 | |
| Unconditional image generation | CelebA unconditional 64 x 64 | FID1.21 | 95 | |
| Classification | CelebA (test) | Average Accuracy95.9 | 92 | |
| Facial Attribute Classification | CelebA (test) | Average Acc92.98 | 89 | |
| Model Inversion Defense | CelebA | Accuracy94.86 | 64 | |
| Image Reconstruction | CelebA HQ (test) | FID (Reconstruction)9.54 | 50 | |
| Image Generation | CelebA (test) | FID3.21 | 49 | |
| Identity Protection | CelebA (test) | ISM77.5 | 48 | |
| Attribute Classification | CelebA (test) | Worst-group Accuracy94.94 | 48 | |
| Image Classification | CelebA-HQ | Accuracy83.87 | 46 | |
| Blind Face Restoration | CelebA (test) | SSIM100 | 44 | |
| Image Inpainting | CelebA-HQ | LPIPS0.009 | 42 | |
| Sparse Recovery | CelebA (test) | MSE0.0247 | 40 | |
| Face Anonymization | CelebA-HQ official (test) | ReID Score24.8 | 40 | |
| Few-shot conditional generation | CelebA-64 (train) | FID3.77 | 40 | |
| Image Generation | CelebA | FID2.78 | 39 | |
| Face Recognition | CelebA (test) | Top-1 Acc96.18 | 39 | |
| Image Classification | CelebA (test) | Accuracy91.9 | 37 | |
| Inpainting | CelebA-HQ | LPIPS0.009 | 36 | |
| Attribute Classification (Blurry) | CelebA | Unbiased Accuracy89.79 | 36 | |
| Attribute Classification (Blond Hair) | CelebA | Unbiased Accuracy92.62 | 36 | |
| Model Inversion | CelebA (test) | Attack Accuracy93.2 | 36 |