Image Classification on CIFAR-10 (Adversarial Error Rates)
31.75Top-1 Error (FGSM, eps=8/255)TENET (A.T.)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TENET (A.T.)Backbone=ResNeXt-29, Adversarial Training=true2021.03 | 31.75 | — | 20.07 | 37.07 | |
| A.T.Backbone=ResNeXt-29, Adversarial Training=true2021.03 | 36.37 | — | 22.61 | 42.82 | |
| TENET (Aug.)Backbone=ResNeXt-29, Data Augmentation=Augmix2021.03 | 60.47 | 3.5 | 90.45 | — | |
| TENETBackbone=ResNeXt-292021.03 | 61.05 | 3.89 | 91.28 | — | |
| CutoutBackbone=ResNeXt-29, Data Augmentation=Cutout2021.03 | 71.07 | 3.97 | 97.19 | — | |
| BaselineBackbone=ResNeXt-292021.03 | 72.81 | 5.72 | 94.15 | — | |
| AugmixBackbone=ResNeXt-29, Data Augmentation=Augmix2021.03 | 76.03 | 3.95 | 93.67 | — |