Image Classification on Caltech-101 (test)
97.5Top-1 Accuracy2SFS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| 2SFSBackbone=ViT-L/14, Shots=162025.03 | 97.5 | — | — | |
| 2SFSBackbone=ViT-B/16, Shots=162025.03 | 96.4 | — | — | |
| 2SFSBackbone=ViT-B/32, Shots=162025.03 | 95.8 | — | — | |
| NAFBackbone=ResNet-502026.05 | 84.41 | — | 96.14 | |
| ERMBackbone=ResNet-502026.05 | 81.99 | — | 94.7 | |
| NAFBackbone=ResNet-182026.05 | 81.94 | — | 95.01 | |
| ERMBackbone=ResNet-182026.05 | 79.2 | — | 93.29 | |
| CLIP-RefineBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 24.79 | — | — | |
| Pre-trainedBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 23.98 | — | — | |
| HyCDBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 23.98 | — | — | |
| Self-KDBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 23.94 | — | — | |
| ContrastiveBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 23.77 | — | — | |
| HyCD + LalignBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 23.73 | — | — | |
| m²-mixBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2025.04 | 23.49 | — | — | |
| AutoAugment-transferBackbone=Inception v4, Augmentation Policy=ImageNet transfer policy, Train Size=3,060, Classes=102, Image Size=448x448, Epochs=1,000, Training Scheme=trained from scratch2018.05 | — | 13.1 | — | |
| BaselineBackbone=Inception v4, Augmentation Policy=Inception pre-processing, Train Size=3,060, Classes=102, Image Size=448x448, Epochs=1,000, Training Scheme=trained from scratch2018.05 | — | 19.4 | — | |
| ResNet-101Setup=G^l = 322023.02 | — | 22.781 | — | |
| ResNet-101Setup=G^l = 12023.02 | — | 26.647 | — | |
| ResNet-101Setup=G^l = G_practical2023.02 | — | 22.247 | — | |
| ResNet-50Setup=G^l = 322023.02 | — | 24.021 | — | |
| ResNet-50Setup=G^l = 12023.02 | — | 24.167 | — | |
| ResNet-50Setup=G^l = G_practical2023.02 | — | 22.368 | — |