Image-to-Text Retrieval on Flickr30k-CN (test)
96.9R@1R2D2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| R2D2Backbone=ViT-L2022.05 | 96.9 | 99.8 | 100 | 96.2 | |
| R2D2Backbone=ViT-L2022.05 | 95.6 | 99.8 | 100 | 95.8 | |
| R2D2Backbone=ViT-B2022.05 | 93.6 | 99.5 | 99.8 | 94.5 | |
| R2D2Backbone=ViT-B2022.05 | 93.2 | 99.2 | 99.8 | 94 | |
| WukongBackbone=ViT-L2022.05 | 92.7 | 99.1 | 99.6 | 93.4 | |
| CLIPBackbone=ViT-L2022.05 | 91.6 | 99.1 | 99.7 | 93.2 | |
| FILIPBackbone=ViT-L2022.05 | 90.6 | 98.8 | 99.6 | 93 | |
| CLIPBackbone=ViT-B2022.05 | 87.1 | 97.7 | 98.8 | 89.7 | |
| WukongBackbone=ViT-B2022.05 | 83.9 | 97.6 | 99 | 88.7 | |
| FILIPBackbone=ViT-B2022.05 | 72.1 | 91.3 | 95.8 | 81.9 |