Referring Expression Retrieval on PhraseCut (test)
64Acc@1MaskInversion
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MaskInversionBackbone=ViT-H/14, zero-shot=true2024.07 | 64 | 96 | 99.2 | |
| MaskInversionBackbone=ViT-L/14, zero-shot=true2024.07 | 60.2 | 94.9 | 98.7 | |
| MaskInversionBackbone=ViT-B/16, zero-shot=true2024.07 | 57.2 | 93.3 | 98.3 | |
| MaskInversionBackbone=ViT-B/32, zero-shot=true2024.07 | 54.8 | 93 | 98.5 | |
| Masked CropBackbone=ViT-B/16, zero-shot=true, Reproduced=true2024.07 | 48.3 | 89.7 | 97.2 | |
| FGVPBackbone=ViT-B/16, zero-shot=true, Reproduced=true2024.07 | 35.9 | 83.5 | 95.2 | |
| AlphaCLIPBackbone=ViT-B/16, zero-shot=false, Reproduced=true2024.07 | 34 | 80 | 93.6 | |
| RedCircleBackbone=ViT-B/16, zero-shot=true, Reproduced=true2024.07 | 21.5 | 72.3 | 90.3 | |
| CropBackbone=ViT-B/16, zero-shot=true, Reproduced=true2024.07 | 15.1 | 67 | 87.6 | |
| CLIPBackbone=ViT-B/16, zero-shot=true, Reproduced=true2024.07 | 14.4 | 66.4 | 87.1 |