Fine-grained Classification on SUN397 (Adversarial)
64.8Clean AccuracySS-TPT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SS-TPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.8 | 45.6 | |
| TPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.6 | 0 | |
| EnsembleBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.1 | 23.3 | |
| TAPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.1 | 13.7 | |
| R-TPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.1 | 43.9 | |
| CLIPBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 62.6 | 0 | |
| DOCBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 58.1 | 7.2 | |
| CLIPBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 33.2 | 6.6 | |
| TPTBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 32.7 | 8.9 | |
| R-TPTBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 27.6 | 13.1 | |
| SS-TPTBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 27.4 | 13.3 | |
| EnsembleBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 26.4 | 12.8 | |
| TTCBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 20.8 | 3.6 |