Image-to-Text Retrieval on Flickr30k short-caption (test)
65.48Recall@1CLIP + TEVI
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLIP + TEVIBackbone=ViT-L/142026.06 | 65.48 | 87.77 | 92.7 | |
| CLIP + TEVIBackbone=ViT-B/162026.06 | 64.2 | 85.7 | 91.12 | |
| CLIPBackbone=ViT-L/142026.06 | 63.71 | 86.49 | 91.32 | |
| SigLIP + TEVIBackbone=ViT-B/162026.06 | 62.82 | 86.19 | 91.62 | |
| SigLIPBackbone=ViT-B/162026.06 | 62.23 | 85.4 | 91.42 | |
| AlignCLIP + TEVIBackbone=ViT-B/162026.06 | 61.24 | 83.53 | 89.15 | |
| SharedCLIPBackbone=ViT-B/162026.06 | 60.75 | 84.22 | 89.64 | |
| CLIPBackbone=ViT-B/162026.06 | 59.66 | 83.73 | 90.14 | |
| SharedCLIP + TEVIBackbone=ViT-B/162026.06 | 59.37 | 83.33 | 89.05 | |
| AlignCLIPBackbone=ViT-B/162026.06 | 57.49 | 82.35 | 89.94 |