Image-to-Text Retrieval on Flickr30k (R@1, R@5, R@10, R@15)
92.5Recall@1FAST-GOAL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FAST-GOALBackbone=ViT-L/142026.05 | 92.5 | 99.4 | 99.7 | 99.8 | |
| EVA-CLIPBackbone=ViT-L/142026.05 | 91.1 | 99.1 | 99.8 | 99.8 | |
| GOALBackbone=ViT-L/142026.05 | 90.8 | 98.8 | 99.5 | 99.7 | |
| Long-CLIPBackbone=ViT-L/142026.05 | 90 | 98.9 | 99.5 | 99.9 | |
| EVA-CLIPBackbone=ViT-B/162026.05 | 87.8 | 98.3 | 99.2 | 99.4 | |
| FAST-GOALBackbone=ViT-B/162026.05 | 86.8 | 97.5 | 98.9 | 99.8 | |
| CLIPBackbone=ViT-L/142026.05 | 86.4 | 97.5 | 99.3 | 99.4 | |
| Long-CLIPBackbone=ViT-B/162026.05 | 85.9 | 98.5 | 99.5 | 99.8 | |
| GOALBackbone=ViT-B/162026.05 | 85.1 | 96.7 | 98.6 | 99.1 | |
| FineCLIPBackbone=ViT-B/162026.05 | 84.6 | 97 | 98.9 | 99.4 | |
| CLIPBackbone=ViT-B/162026.05 | 82.9 | 97.2 | 98.9 | 99.4 | |
| FineCLIPBackbone=ViT-L/142026.05 | 80.8 | 95.9 | 97.6 | 98.7 |