Image Classification on ImageNet 16-shot (Source)
73.2AccuracyCross-Modal Linear Probing
Evaluation results
| Method | Links | |
|---|---|---|
| Cross-Modal Linear ProbingBackbone=ViT-B/162023.01 | 73.2 | |
| WiSE-FTBackbone=ViT-B/16, Interpolation coefficient (alpha)=0.52023.01 | 73 | |
| Cross-Modal WiSE-FTBackbone=ViT-B/16, Interpolation coefficient (alpha)=0.52023.01 | 72.9 | |
| CoOpBackbone=ViT-B/16, Prompt length (M)=42023.01 | 71.9 | |
| CoOpBackbone=ViT-B/16, Prompt length (M)=162023.01 | 71.7 | |
| NOAHShots per category=162022.06 | 71.5 | |
| CoCoOpBackbone=ViT-B/162023.01 | 71 | |
| LoRAShots per category=162022.06 | 70.8 | |
| AdapterShots per category=162022.06 | 70.5 | |
| VPTShots per category=162022.06 | 70.5 | |
| Zero-Shot CLIPBackbone=ViT-B/16, Evaluation Protocol=Zero-shot2023.01 | 66.7 | |
| Linear ProbingBackbone=ViT-B/16, Evaluation Protocol=Linear Probing2023.01 | 65.9 | |
| Cross-Modal WiSE-FTBackbone=ResNet50, Interpolation coefficient (alpha)=0.52023.01 | 65.2 | |
| Cross-Modal Linear ProbingBackbone=ResNet502023.01 | 64.5 | |
| CoOpBackbone=ResNet50, Prompt length (M)=162023.01 | 63.3 | |
| CoOpBackbone=ResNet50, Prompt length (M)=42023.01 | 63 | |
| WiSE-FTBackbone=ResNet50, Interpolation coefficient (alpha)=0.52023.01 | 62.9 | |
| Zero-Shot CLIPBackbone=ResNet50, Evaluation Protocol=Zero-shot2023.01 | 58.2 | |
| Linear ProbingBackbone=ResNet50, Evaluation Protocol=Linear Probing2023.01 | 55.9 |