Image-to-Text Retrieval on MS COCO short-caption (test)
37.88R@1CLIP + TEVI
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLIP + TEVIBackbone=ViT-L/142026.06 | 37.88 | 64.38 | 75.16 | |
| CLIPBackbone=ViT-L/142026.06 | 36.52 | 62.5 | 73.06 | |
| SigLIP + TEVIBackbone=ViT-B/162026.06 | 36.24 | 61.2 | 72.82 | |
| CLIP + TEVIBackbone=ViT-B/162026.06 | 35.66 | 61.82 | 72.94 | |
| AlignCLIP + TEVIBackbone=ViT-B/162026.06 | 34.42 | 60.3 | 71.48 | |
| SigLIPBackbone=ViT-B/162026.06 | 33.88 | 60.3 | 70.62 | |
| SharedCLIP + TEVIBackbone=ViT-B/162026.06 | 33.82 | 59.86 | 70.9 | |
| CLIPBackbone=ViT-B/162026.06 | 32.98 | 59.02 | 70.04 | |
| AlignCLIPBackbone=ViT-B/162026.06 | 32.7 | 58.92 | 70.46 | |
| SharedCLIPBackbone=ViT-B/162026.06 | 32.62 | 58.88 | 69.94 |