AI-generated image detection on AIGIBench PhotoMaker
97AccuracySAFE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SAFETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 97 | 99.3 | |
| PGCTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 93.7 | 98.6 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 92.7 | 95.2 | |
| AIDETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 91.7 | 95.6 | |
| AIDEevaluation_strategy=cross-model2025.01 | 83.4 | 80.9 | |
| DDATraining Data=ProGAN (4 classes) and SDv1.42026.05 | 78.7 | 90.8 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 77 | 74.9 | |
| FreqNetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 77 | 74.9 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 69.3 | 66.6 | |
| UnivFDTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 69.3 | 72.3 | |
| LEGOVenue=-2026.05 | 62.2 | 80.5 | |
| B-FreeTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 61.6 | 74.8 | |
| LGradTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 59.9 | 67.2 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 50.6 | 52 | |
| CNNDetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 50.1 | 58.2 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 49.6 | 51.5 | |
| NPRTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 48.1 | 43.6 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 47 | 41.4 | |
| NPRevaluation_strategy=cross-model2025.01 | 46.4 | 52 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 46.4 | 37.4 | |
| LGradevaluation_strategy=cross-model2025.01 | 32 | 35.8 |