Image Forgery Detection on AIGCDetectionBenchmark (test)
100Detection Score (ProGAN)CNNSpot
Evaluation results
| Method | Links | |||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CNNSpotEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 100 | 90.17 | 71.17 | 87.62 | 94.6 | 81.42 | 86.91 | 91.65 | 60.39 | 58.07 | 51.39 | 50.57 | 50.53 | 56.46 | 51.03 | 50.45 | 70.78 | — | — | |
| FusingEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 100 | 85.2 | 77.4 | 87 | 97 | 77 | 83.3 | 66.8 | 49 | 57.2 | 52.2 | 51 | 51.4 | 55.1 | 51.7 | 52.8 | 68.38 | — | — | |
| PatchCraftEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 100 | 92.77 | 95.8 | 70.17 | 99.97 | 71.58 | 89.55 | 85.8 | 82.17 | 83.79 | 90.12 | 95.38 | 95.3 | 88.91 | 91.07 | 96.6 | 89.31 | — | — | |
| AIDEEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 99.99 | 99.64 | 83.95 | 98.48 | 99.91 | 73.25 | 98 | 94.2 | 93.43 | 95.09 | 77.2 | 93 | 92.85 | 95.16 | 93.55 | 96.6 | 92.77 | — | — | |
| LGradEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 99.83 | 91.08 | 85.62 | 86.94 | 99.27 | 78.46 | 85.32 | 55.7 | 67.15 | 66.11 | 65.35 | 63.02 | 63.67 | 72.99 | 59.55 | 65.45 | 75.34 | — | — | |
| UniverFDEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 99.81 | 84.93 | 95.08 | 98.33 | 95.75 | 99.47 | 74.96 | 86.9 | 66.87 | 62.46 | 56.13 | 63.66 | 63.49 | 85.31 | 70.93 | 50.75 | 78.43 | — | — | |
| NPREvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 99.79 | 97.7 | 84.35 | 96.1 | 99.35 | 82.5 | 98.38 | 65.8 | 69.69 | 78.36 | 77.85 | 78.63 | 78.89 | 78.13 | 76.11 | 64.9 | 82.91 | — | — | |
| LNPEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 99.67 | 91.75 | 77.75 | 84.1 | 99.92 | 75.39 | 94.64 | 70.85 | 84.73 | 80.52 | 65.55 | 85.55 | 85.67 | 74.46 | 82.06 | 88.75 | 83.84 | — | — | |
| FGTS (Linear probe)Evaluation Protocol=Linear probe, Backbone=Frozen DINOv2, Reference Set=1k real / 1k fake2025.11 | 99.59 | 89.18 | 99.18 | 89.53 | 94.52 | 99.8 | 90.92 | 98.2 | 85.76 | 93.04 | 77.32 | 93.52 | 92.66 | 98.72 | 95.6 | 77 | 92.45 | — | — | |
| FreDectEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 99.36 | 78.02 | 81.97 | 78.77 | 94.62 | 80.57 | 66.19 | 50.75 | 63.42 | 54.13 | 45.87 | 38.79 | 39.21 | 77.8 | 40.3 | 34.7 | 64.03 | — | — | |
| FGTS (Training free)Evaluation Protocol=Training-free, Backbone=Frozen DINOv22025.11 | 99.25 | 77.88 | 75.73 | 69.84 | 50.1 | 99.79 | 75.71 | 77.2 | 59.9 | 75.1 | 50.6 | 74.9 | 72.9 | 94.4 | 80 | 60.5 | 78.99 | — | — | |
| DIRE-GEvaluation Protocol=Fine-tuned, Training Data Configuration=ProGAN2025.11 | 95.19 | 83.03 | 70.12 | 74.19 | 95.47 | 67.79 | 75.31 | 58.05 | 75.78 | 71.75 | 58.01 | 49.74 | 49.83 | 53.68 | 54.46 | 66.48 | 68.68 | — | — | |
| DIRE-DEvaluation Protocol=Official setup, Training Data Configuration=ADM2025.11 | 52.75 | 51.31 | 49.7 | 49.58 | 46.72 | 51.23 | 51.72 | 53.3 | 98.25 | 92.42 | 89.45 | 91.24 | 91.63 | 91.9 | 90.9 | 92.45 | 71.53 | — | — | |
| CLIP2025.10 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 65.9 | 64.7 | |
| FakeShield2025.10 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 78.1 | 75.4 | |
| HiFi-Net2025.10 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 67.8 | 66.4 | |
| ResNet2025.10 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 76.5 | 77.4 | |
| UniShield2025.10 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 94.2 | 93.1 |