Text-to-Image Retrieval on COCO-CN (test)
79.1R@1R2D2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| R2D2Backbone=ViT-L2022.05 | 79.1 | 96.5 | 98.9 | — | |
| R2D2Backbone=ViT-B2022.05 | 76 | 94.9 | 98.3 | — | |
| WukongBackbone=ViT-L2022.05 | 74 | 94.4 | 98.1 | — | |
| FILIPBackbone=ViT-L2022.05 | 72.2 | 92.4 | 97.2 | — | |
| CLIPBackbone=ViT-L2022.05 | 70.1 | 92.2 | 96.4 | — | |
| CLIPBackbone=ViT-B2022.05 | 68.9 | 93.3 | 97.3 | — | |
| WukongBackbone=ViT-B2022.05 | 67 | 91.4 | 96.7 | — | |
| FILIPBackbone=ViT-B2022.05 | 56.2 | 86.8 | 94.3 | — |