Image Classification on ImageNet-1k (test) (Zero-shot)
37.6Zero-shot AccuracyCLIP + MaskSub
Evaluation results
| Method | Links | |
|---|---|---|
| CLIP + MaskSubArchitecture=ViT-B/32, Training strategy=CLIP with MaskSub2023.06 | 37.6 | |
| CLIPArchitecture=ViT-B/32, Training strategy=Standard CLIP2023.06 | 33.5 | |
| CLIP + MaskingArchitecture=ViT-B/32, Training strategy=CLIP with standard masking2023.06 | 29.8 |