Fine-grained classification on Describable Textures Dataset (DTD)
55.4AccuracyC-TPT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| C-TPTPre-trained Backbone=CLIP-ViT-L/14, Perturbation budget (epsilon)=4.02025.04 | 55.4 | 0 | |
| TPTPre-trained Backbone=CLIP-ViT-L/14, Perturbation budget (epsilon)=4.02025.04 | 55.1 | 0 | |
| R-TPTPre-trained Backbone=CLIP-ViT-L/14, Perturbation budget (epsilon)=4.02025.04 | 54 | 38 | |
| MTAPre-trained Backbone=CLIP-ViT-L/14, Perturbation budget (epsilon)=4.02025.04 | 53.4 | 27.2 | |
| CLIPPre-trained Backbone=CLIP-ViT-L/14, Perturbation budget (epsilon)=4.02025.04 | 52.4 | 0 | |
| EnsemblePre-trained Backbone=CLIP-ViT-L/14, Perturbation budget (epsilon)=4.02025.04 | 51.3 | 31.3 | |
| APMPre-trained on ImageNet=false2024.10 | 48.9 | — | |
| TPTPre-trained on ImageNet=false2024.10 | 47.8 | — | |
| MTABackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 46.5 | 16.2 | |
| R-TPTBackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 46.4 | 32.8 | |
| CoCoOpPre-trained on ImageNet=true2024.10 | 45.5 | — | |
| EnsemblePre-trained on ImageNet=false2024.10 | 45 | — | |
| CLIPBackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 44.4 | 0 | |
| CLIP-VIT-B/16Pre-trained on ImageNet=false2024.10 | 44.3 | — | |
| TTPBackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 44.1 | 36 | |
| MTABackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 43.8 | 28.8 | |
| AGCBackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 43.4 | 88.4 | |
| CLIPPre-trained Backbone=ViT-B/32, Perturbation budget (epsilon)=4.02025.04 | 43 | 0 | |
| CLIPBackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 43 | 0 | |
| TTPBackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 42.8 | 32.2 | |
| R-TPTPre-trained Backbone=ViT-B/32, Perturbation budget (epsilon)=4.02025.04 | 42.1 | 29.1 | |
| R-TPTBackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 42.1 | 29.1 | |
| CoOpPre-trained on ImageNet=true2024.10 | 41.9 | — | |
| EnsembleBackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 41.7 | 31.9 | |
| R-TPTPre-trained Backbone=ResNet50, Perturbation budget (epsilon)=1.02025.04 | 41.3 | 33.5 | |
| AGCBackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 41.2 | 71.2 | |
| TTCBackbone=ViT-B/16, Epsilon (ε)=4/2552026.05 | 41 | 4.5 | |
| CLIPPre-trained Backbone=ResNet50, Perturbation budget (epsilon)=1.02025.04 | 40.4 | 0.8 | |
| EnsembleBackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 39.6 | 30.1 | |
| APT+TeCoAPre-trained Backbone=ResNet50, Perturbation budget (epsilon)=1.02025.04 | 39.2 | 39 | |
| TTCBackbone=ViT-B/32, Epsilon (ε)=4/2552026.05 | 37.3 | 4.7 | |
| APT+TeCoAPre-trained Backbone=ViT-B/32, Perturbation budget (epsilon)=4.02025.04 | 35.2 | 33.7 | |
| FAREPre-trained Backbone=ViT-B/32, Perturbation budget (epsilon)=4.02025.04 | 28.3 | 27.3 | |
| TeCoAPre-trained Backbone=ResNet50, Perturbation budget (epsilon)=1.02025.04 | 26.2 | 26 | |
| TeCoAPre-trained Backbone=ViT-B/32, Perturbation budget (epsilon)=4.02025.04 | 24.5 | 24 | |
| APTPre-trained Backbone=ResNet50, Perturbation budget (epsilon)=1.02025.04 | 16.6 | 7.9 | |
| APTPre-trained Backbone=ViT-B/32, Perturbation budget (epsilon)=4.02025.04 | 9 | 0.1 |