List your company
ResearchFormerly WizWand

Image Classification Robustness on CIFAR-100 (test) (Clean and AutoAttack)

65Clean AccuracyAT(TrH)
58.34460.07261.863.528Nov 22, 2022
Updated 8mo ago

Evaluation results

MethodLinks
AT(TrH)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.116533
AT(SWA)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1164.133.7
TRADES(TrH)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1163.833
TRADES(S2O)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1163.531.5
AT(S2O)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1163.231.5
TRADES(SWA)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1162.733
AT(AWP)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1161.532.2
TRADES(AWP)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1160.832.2
AT(base)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.116030.4
TRADES(base)Backbone=ViT-B16, Perturbation budget (epsilon)=8/255, Adversarial threat model=l_inf2022.1158.630.5