Adversarial Robust Image Classification on CIFAR-10 (test) (Attack Specific Statistics)
94.57Clean Accuracy AvgNatural
Evaluation results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NaturalBackbone=ResNet-182023.12 | 94.57 | 86.3 | 0.035 | 18.6 | 9 | 0.436 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
| BATBackbone=ResNet-182023.12 | 86.72 | 72.3 | 0.092 | 60.97 | 33.8 | 0.255 | 49.6 | 22.7 | 0.325 | 47.49 | 19.5 | 0.354 | 48.18 | 20.7 | 0.341 | |
| AdaADBackbone=ResNet-182023.12 | 84.73 | 68.1 | 0.114 | 59.7 | 34.8 | 0.246 | 53.82 | 29.3 | 0.285 | 52.3 | 26 | 0.312 | 50.91 | 24.7 | 0.322 | |
| SATBackbone=ResNet-182023.12 | 84.03 | 63.9 | 0.118 | 56.71 | 26.7 | 0.283 | 49.34 | 21 | 0.332 | 48.99 | 19.9 | 0.352 | 46.44 | 16.8 | 0.385 | |
| Fair-ARDBackbone=ResNet-182023.12 | 83.81 | 69.4 | 0.112 | 58.41 | 38.5 | 0.251 | 50.91 | 29.9 | 0.296 | 49.96 | 28.3 | 0.312 | 47.97 | 25.1 | 0.338 | |
| ABSLDBackbone=ResNet-182023.12 | 83.04 | 68.1 | 0.103 | 59.83 | 40.5 | 0.202 | 54.5 | 36.5 | 0.216 | 51.77 | 32.8 | 0.249 | 50.25 | 31 | 0.256 | |
| RSLADBackbone=ResNet-182023.12 | 82.94 | 66.3 | 0.122 | 59.51 | 34.7 | 0.244 | 54 | 28.5 | 0.276 | 52.51 | 27 | 0.296 | 51.25 | 25.5 | 0.304 | |
| FRLBackbone=ResNet-182023.12 | 82.29 | 64.6 | 0.114 | 55.03 | 37.1 | 0.23 | 49.05 | 31.7 | 0.248 | 47.88 | 30.4 | 0.266 | 46.54 | 28.1 | 0.28 | |
| TRADESBackbone=ResNet-182023.12 | 81.45 | 67.6 | 0.113 | 56.65 | 26.7 | 0.267 | 51.78 | 30.4 | 0.301 | 49.15 | 27.1 | 0.341 | 48.17 | 25.9 | 0.35 | |
| CFABackbone=ResNet-182023.12 | 78.64 | 63.6 | 0.123 | 57.95 | 36.8 | 0.231 | 54.42 | 33.3 | 0.258 | 50.91 | 27.5 | 0.288 | 50.37 | 26.7 | 0.296 |