AI-generated image detection on AIGIBench BlendFace
77.9AccuracyDDA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DDATraining Data=ProGAN (4 classes) and SDv1.42026.05 | 77.9 | 90 | |
| PGCTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 71.8 | 82.3 | |
| B-FreeTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 60.3 | 79.9 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 55.2 | 60.5 | |
| CNNDetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 52.4 | 73.4 | |
| LEGOVenue=-2026.05 | 51.9 | 59.4 | |
| AIDETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 51.5 | 54.2 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 49.3 | 34.2 | |
| SAFETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 47.3 | 45.6 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 47.2 | 42.7 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 46 | 44.7 | |
| NPRTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 44.5 | 34.7 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 44.1 | 44 | |
| AIDEevaluation_strategy=cross-model2025.01 | 43.4 | 43.8 | |
| LGradTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 41.8 | 34.9 | |
| LGradevaluation_strategy=cross-model2025.01 | 38.6 | 39.4 | |
| UnivFDTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 35 | 35.3 | |
| NPRevaluation_strategy=cross-model2025.01 | 33.9 | 33.3 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 28.2 | 38 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 23.3 | 34.1 | |
| FreqNetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 23.3 | 34.1 |