Adversarial Image Classification on CIFAR-10 (test) (PGD-40)
88.9Clean AccuracyDAJAT (Base, 2*AA)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DAJAT (Base, 2*AA)Backbone=WideResNet-34-10, Training Epochs=1102022.10 | 88.9 | 60.97 | |
| Rebuffi et al.Backbone=WideResNet-34-10, Training Epochs=110, Augmentation=CutMix2022.10 | 87.24 | 57.6 | |
| TRADES-AWPBackbone=WideResNet-34-10, Training Epochs=1102022.10 | 85.35 | 59.13 | |
| TRADESBackbone=WideResNet-34-10, Training Epochs=1102022.10 | 84.72 | 56.92 |