Adversarial Fine-grained Classification on Cars
67.7Clean AccuracyMTA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MTABackbone=CLIP-ViT-B/16, Perturbation bound (epsilon)=4.02025.12 | 67.7 | 18.5 | |
| R-TPTBackbone=CLIP-ViT-B/16, Perturbation bound (epsilon)=4.02025.12 | 67 | 34.7 | |
| TPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 66.3 | 0 | |
| EnsembleBackbone=CLIP-ViT-B/16, Perturbation bound (epsilon)=4.02025.12 | 65.7 | 26 | |
| CLIPBackbone=CLIP-ViT-B/16, Perturbation bound (epsilon)=4.02025.12 | 65.5 | 0 | |
| CLIPBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 65.5 | 0 | |
| TTPBackbone=CLIP-ViT-B/16, Perturbation bound (epsilon)=4.02025.12 | 65.4 | 37.4 | |
| SS-TPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 65.1 | 33.6 | |
| EnsembleBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.8 | 10.3 | |
| R-TPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 64.5 | 30.6 | |
| MTABackbone=CLIP-ViT-B/32, Perturbation Bound (epsilon)=4.02025.12 | 63.4 | 26.4 | |
| TAPTBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 63.2 | 4.4 | |
| R-TPTBackbone=CLIP-ViT-B/32, Perturbation Bound (epsilon)=4.02025.12 | 63.1 | 28.4 | |
| DOCBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 62 | 2.2 | |
| TTCBackbone=CLIP-ViT-B/16, Augmented views=15, Perturbation (epsilon)=4.02026.06 | 60.9 | 3.3 | |
| CLIPBackbone=CLIP-ViT-B/32, Perturbation Bound (epsilon)=4.02025.12 | 60.1 | 0 | |
| TTPBackbone=CLIP-ViT-B/32, Perturbation Bound (epsilon)=4.02025.12 | 59.8 | 29.8 | |
| MTABackbone=CLIP-ResNet50, Epsilon=1.02025.04 | 58.7 | 17.8 | |
| TPTBackbone=CLIP-ResNet50, Epsilon=1.02025.04 | 58.4 | 0 | |
| R-TPTBackbone=CLIP-ResNet50, Epsilon=1.02025.04 | 58.1 | 42.9 | |
| EnsembleBackbone=CLIP-ResNet50, Epsilon=1.02025.04 | 57.1 | 36.2 | |
| C-TPTBackbone=CLIP-ResNet50, Epsilon=1.02025.04 | 56.6 | 0 | |
| CLIPBackbone=CLIP-ResNet50, Epsilon=1.02025.04 | 55.7 | 0 | |
| TTCBackbone=CLIP-ViT-B/16, Perturbation bound (epsilon)=4.02025.12 | 55 | 2.9 | |
| EnsembleBackbone=CLIP-ViT-B/32, Perturbation Bound (epsilon)=4.02025.12 | 51.7 | 25.9 | |
| TTCBackbone=CLIP-ViT-B/32, Perturbation Bound (epsilon)=4.02025.12 | 48.1 | 2.3 | |
| CLIPBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 10.2 | 1.1 | |
| TPTBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 9.6 | 2 | |
| SS-TPTBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 6.4 | 3.1 | |
| R-TPTBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 6.1 | 2.9 | |
| EnsembleBackbone=CLIP-ViT-B/32, Pre-trained=TeCoA, Augmented views=15, Epsilon=4.02026.06 | 5.8 | 2.7 |