AI-generated image detection on AIGIBench SD3
97.4AccuracyLEGO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LEGOVenue=-2026.05 | 97.4 | 99.8 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 95.4 | 97.6 | |
| AIDEevaluation_strategy=cross-model2025.01 | 89.7 | 94.1 | |
| NPRevaluation_strategy=cross-model2025.01 | 86.4 | 93.5 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 83.6 | 87.6 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 77.6 | 84 | |
| LGradevaluation_strategy=cross-model2025.01 | 66.2 | 71.4 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 62.9 | 68.5 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 53.5 | 65.6 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 52.8 | 66.3 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 49 | 48.6 |