Image Classification on CIFAR-10 (test) (Natural and Robust Accuracy)
96.93Natural AccuracyShi et al.
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Shi et al.Architecture=WRN-28-102023.09 | 96.93 | 63.1 | |
| Raw WideResNetArchitecture=WRN-28-102023.09 | 95.8 | 0 | |
| Yoon et al.Architecture=WRN-28-10, sigma=0.12023.09 | 93.09 | 85.45 | |
| LGAPArchitecture=WRN-28-102023.09 | 90.03 | 71.68 | |
| Song et al.Architecture=ResNet-62, Purification=AT + PixelCNN2023.09 | 90 | 70 | |
| Madry et al.Architecture=ResNet-562023.09 | 87.3 | 70.2 | |
| Yoon et al.Architecture=WRN-28-10, sigma=0.252023.09 | 86.14 | 80.24 | |
| Dong et al.Architecture=ResNet-182023.09 | 84.98 | 51.29 | |
| Hill et al.Architecture=WRN-28-102023.09 | 84.12 | 78.91 | |
| Song et al.Architecture=ResNet-62, Purification=Natural + PixelCNN2023.09 | 82 | 61 | |
| Grathwohl et al.Architecture=WRN-28-102023.09 | 75.5 | 23.8 | |
| Du et al.Architecture=WRN-28-102023.09 | 48.7 | 37.5 |