AI-generated image detection on AIGIBench SDXL
97.6AccuracyLEGO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LEGOVenue=-2026.05 | 97.6 | 99.8 | |
| DFFreqevaluation_strategy=cross-model2025.01 | 95.9 | 98.7 | |
| AIDEevaluation_strategy=cross-model2025.01 | 88.9 | 90 | |
| NPRevaluation_strategy=cross-model2025.01 | 87 | 93.7 | |
| FreqNetevaluation_strategy=cross-model2025.01 | 86.4 | 93.6 | |
| LaDeDaevaluation_strategy=cross-model2025.01 | 83.1 | 88 | |
| FatFormerevaluation_strategy=cross-model2025.01 | 75 | 83 | |
| LGradevaluation_strategy=cross-model2025.01 | 64 | 66.3 | |
| C2P-CLIPevaluation_strategy=cross-model2025.01 | 59.3 | 76 | |
| CNN-Spotevaluation_strategy=cross-model2025.01 | 52.6 | 65.5 | |
| UnivFDevaluation_strategy=cross-model2025.01 | 50.7 | 53.6 |