Text-to-Image Retrieval on MS COCO short-caption (test)
25.04Recall@1CLIP + TEVI
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CLIP + TEVIBackbone=ViT-L/142026.06 | 25.04 | 49.74 | 61.35 | |
| CLIPBackbone=ViT-L/142026.06 | 23.85 | 48.42 | 59.55 | |
| CLIP + TEVIBackbone=ViT-B/162026.06 | 23.12 | 47.03 | 58.67 | |
| SigLIP + TEVIBackbone=ViT-B/162026.06 | 22.37 | 46.6 | 58.66 | |
| SharedCLIP + TEVIBackbone=ViT-B/162026.06 | 22.13 | 45.95 | 57.38 | |
| SigLIPBackbone=ViT-B/162026.06 | 22.06 | 46.21 | 57.7 | |
| AlignCLIPBackbone=ViT-B/162026.06 | 21.79 | 44.55 | 56.41 | |
| SharedCLIPBackbone=ViT-B/162026.06 | 21.54 | 44.81 | 56.87 | |
| CLIPBackbone=ViT-B/162026.06 | 21.38 | 45.3 | 57.03 | |
| AlignCLIP + TEVIBackbone=ViT-B/162026.06 | 21.35 | 44.39 | 55.72 |