Image Classification on UCF101 (Clean and Adversarial Accuracy)
75Clean Top-1 AccuracyTTC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TTCBackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 75 | 2.2 | |
| MTABackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 74.7 | 47.5 | |
| R-TPTBackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 74.3 | 55.6 | |
| CLIPBackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 73.7 | 0 | |
| TTPBackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 73.6 | 57.4 | |
| T-VSSBackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 73.5 | 55.9 | |
| EnsembleBackbone=ViT-L/14, Perturbation budget (epsilon)=4/2552026.06 | 71.7 | 48.3 | |
| MTABackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 67.5 | 27.5 | |
| R-TPTBackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 67.2 | 43.2 | |
| T-VSSBackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 66 | 45.2 | |
| TTCBackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 65.8 | 1.6 | |
| CLIPBackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 65.2 | 0 | |
| TTPBackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 65 | 47.2 | |
| EnsembleBackbone=ViT-B/16, Perturbation budget (epsilon)=4/2552026.06 | 63 | 30.6 | |
| TPTBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 60.6 | 0.3 | |
| MTABackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 60.6 | 31.3 | |
| C-TPTBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 60.1 | 0.1 | |
| R-TPTBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 59.7 | 50.9 | |
| CLIPBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 58.9 | 0 | |
| T-VSSBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 58.9 | 52.5 | |
| TTPBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 58.5 | 49 | |
| EnsembleBackbone=ResNet-50, Perturbation budget (epsilon)=1/2552026.06 | 53.9 | 43 | |
| C-TPTBackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 36.4 | 8.1 | |
| TPTBackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 34.9 | 10.2 | |
| CLIP-TeCoABackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 34.6 | 6.7 | |
| MTABackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 34.6 | 12.5 | |
| T-VSSBackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 31.5 | 19.6 | |
| R-TPTBackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 31.1 | 17.4 | |
| EnsembleBackbone=TeCoA-pretrained CLIP-ViT-B/32, Epsilon (ϵ)=4/2552026.06 | 26.4 | 14 |