AI-generated image detection on GenImage robustness evaluation
85.64Accuracy (JPEG QF=90)SARE
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SARE2025.08 | 85.64 | 82.72 | 79.14 | 87.6 | 82.74 | |
| DRCT2025.08 | 80.97 | 78.06 | 76.18 | 79.51 | 74.75 | |
| UnivFD2025.08 | 73.24 | 70.42 | 69.75 | 75.96 | 73.99 | |
| Conv-B2025.08 | 71.91 | 71.57 | 71.42 | 75.3 | 75.28 | |
| GramNet2025.08 | 71.22 | 71.02 | 71.22 | 69.89 | 66.31 | |
| NPR2025.08 | 71.1 | 70.71 | 70.07 | 72.22 | 71.54 | |
| DE-FAKE2025.08 | 71 | 70.88 | 70.44 | 72.33 | 71.88 | |
| AIDE2025.08 | 57.22 | 58.5 | 60.45 | 83.13 | 83.79 | |
| DIRE2025.08 | 52.79 | 50.65 | 50.34 | 51.94 | 53.77 |