Image Classification on CLIP-based Base-to-Novel Generalization Benchmark (test)
84.34Base AccuracyZero-init.
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Zero-init.Backbone=CLIP, Mechanism=Zero-initialized Attention2023.03 | 84.34 | 72.84 | 78.13 | |
| CLIPBackbone=CLIP2023.03 | 82.69 | 63.22 | 71.66 | |
| CoOpBackbone=CLIP2023.03 | 82.69 | 63.22 | 71.66 | |
| MaPLeBackbone=CLIP2023.03 | 82.28 | 70.14 | 75.73 | |
| CoCoOpBackbone=CLIP2023.03 | 80.47 | 71.69 | 75.83 |