Object Classification on ImageNet (test)
82.8Top-1 AccGLA
Evaluation Results
| Method | Links | |
|---|---|---|
| GLAbackbone=ViT-B/162023.10 | 82.8 | |
| WiSE-FTbackbone=ViT-B/162023.10 | 81.7 | |
| End-to-end fine-tuningbackbone=ViT-B/162023.10 | 81.3 | |
| Linear probebackbone=ViT-B/162023.10 | 79.9 | |
| GLAbackbone=ViT-B/322023.10 | 77.9 | |
| WiSE-FTbackbone=ViT-B/322023.10 | 76.9 | |
| End-to-end fine-tuningbackbone=ViT-B/322023.10 | 76.3 | |
| Linear probebackbone=ViT-B/322023.10 | 75.8 | |
| ZS_EnEvaluation Protocol=Zero-shot2023.11 | 70.66 | |
| Zero-shotbackbone=ViT-B/162023.10 | 68.3 | |
| CLIPBackbone=ViT-B/16, Evaluation Protocol=Zero-shot2023.11 | 66.72 | |
| Zero-shotbackbone=ViT-B/322023.10 | 63.2 | |
| CLIPBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2023.11 | 62.04 | |
| CLIPBackbone=RN101, Evaluation Protocol=Zero-shot2023.11 | 61.26 | |
| CLIPBackbone=RN50, Evaluation Protocol=Zero-shot2023.11 | 58.23 |