AI-generated image detection on AIGIBench InSwap
82AccuracyDDA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DDATraining Data=ProGAN (4 classes) and SDv1.42026.05 | 82 | 87.8 | |
| PGCTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 80.4 | 92.3 | |
| B-FreeTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 62.1 | 79.6 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 59.5 | 69 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 57.9 | 66.4 | |
| LEGOVenue=-2026.05 | 55.4 | 65.5 | |
| CNNDetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 54.5 | 77.9 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 52.1 | 63.7 | |
| AIDETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 50.9 | 54.6 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 50.1 | 44.3 | |
| SAFETraining Data=ProGAN (4 classes) and SDv1.42026.05 | 49.7 | 49.9 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 48.9 | 48.2 | |
| NPRTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 47.8 | 40.7 | |
| AIDEevaluation_strategy=cross-model2025.01 | 46.1 | 43 | |
| LGradTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 44.6 | 35 | |
| UnivFDTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 43.7 | 40.2 | |
| NPRevaluation_strategy=cross-model2025.01 | 41 | 36.9 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 38.3 | 41.8 | |
| LGradevaluation_strategy=cross-model2025.01 | 37.7 | 38.5 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 37.5 | 42.1 | |
| FreqNetTraining Data=ProGAN (4 classes) and SDv1.42026.05 | 37.5 | 42.1 |