Fine-grained Image Classification on DTD
54Top-1 AccuracyR-TPT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| R-TPTCategory=Tuning-based, Backbone=CLIP-ViT-L/14, Attack Protocol=PGD-100, eps=4.02026.06 | 54 | 38 | |
| MTACategory=Tuning-free, Backbone=CLIP-ViT-L/14, Attack Protocol=PGD-100, eps=4.02026.06 | 53.4 | 27.2 | |
| CLIPCategory=Baseline, Backbone=CLIP-ViT-L/14, Attack Protocol=PGD-100, eps=4.02026.06 | 52.4 | 0 | |
| MACCategory=Tuning-free, Backbone=CLIP-ViT-L/14, Attack Protocol=PGD-100, eps=4.02026.06 | 52.1 | 45.3 | |
| TTCCategory=Tuning-free, Backbone=CLIP-ViT-L/14, Attack Protocol=PGD-100, eps=4.02026.06 | 51 | 5.3 | |
| TPTmode=Test-Time Adaptation2026.02 | 47.04 | — | |
| MTAmode=Test-Time Adaptation2026.02 | 46.28 | — | |
| R-TPTmode=Test-Time Adaptation2026.02 | 46.16 | — | |
| RLCFmode=Test-Time Adaptation2026.02 | 46.04 | — | |
| TPSmode=Test-Time Adaptation2026.02 | 46.04 | — | |
| TTLmode=Test-Time Adaptation2026.02 | 45.86 | — | |
| Fair Context Learning (FCL)mode=Test-Time Adaptation2026.02 | 45.45 | — | |
| ZEROmode=Test-Time Adaptation2026.02 | 45.39 | — | |
| C-TPTmode=Test-Time Adaptation2026.02 | 45.27 | — | |
| CLIPmode=Zero-shot2026.02 | 44.39 | — | |
| R-TPTBackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 41.3 | 33.5 | |
| TTPBackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 41 | 33.3 | |
| CLIPBackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 40.4 | 0.8 | |
| APT+TeCoABackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 39.2 | 39 | |
| AGCBackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 39.1 | 53.4 | |
| TeCoABackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 26.2 | 26 | |
| APTBackbone=ResNet50, Attack Type=PGD, Epsilon (ε)=1.02026.05 | 16.6 | 7.9 |