Base-to-new Generalization on Average
85.48Base ScorePromptSRC + T_En
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PromptSRC + T_EnBase Models=ViT-B/32, ViT-B/162023.11 | 85.48 | 77.17 | 81.11 | |
| PromptSRCVision Encoder=ViT-B/162023.11 | 84.26 | 76.1 | 79.97 | |
| PromptSRCVision Encoder=ViT-B/16, reproduced=true2023.11 | 83.81 | 75.69 | 79.54 | |
| Tip-Adapter + SHIPVision Encoder=ViT-B/162023.11 | 83.8 | 76.42 | 79.94 | |
| CoCoOp + T_EnBase Models=RN50, RN101, ViT-B/32, ViT-B/162023.11 | 83.56 | 75.27 | 79.2 | |
| CoOpVision Encoder=ViT-B/162023.11 | 82.69 | 63.22 | 71.66 | |
| MaPLeVision Encoder=ViT-B/162023.11 | 82.28 | 75.14 | 78.55 | |
| ProDAVision Encoder=ViT-B/162023.11 | 81.56 | 72.3 | 76.65 | |
| CoCoOpVision Encoder=ViT-B/162023.11 | 80.47 | 71.69 | 75.83 | |
| CoCoOpVision Encoder=ViT-B/16, reproduced=true2023.11 | 80.14 | 71.55 | 75.6 | |
| T_EnVision Encoder=ViT-B/162023.11 | 72.56 | 77.72 | 75.05 | |
| TF_EnVision Encoder=ViT-B/162023.11 | 72.12 | 77.37 | 74.65 | |
| CLIPVision Encoder=ViT-B/162023.11 | 69.34 | 74.22 | 71.7 |