Text-to-Image Retrieval on Flickr30k short-caption (test)
47.14Recall@1CLIP + TEVI
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLIP + TEVIBackbone=ViT-L/142026.06 | 47.14 | 74.58 | 82.5 | |
| CLIPBackbone=ViT-L/142026.06 | 46.51 | 72.84 | 81.66 | |
| CLIP + TEVIBackbone=ViT-B/162026.06 | 44.75 | 72.1 | 81.22 | |
| SigLIPBackbone=ViT-B/162026.06 | 44.69 | 70.61 | 79.33 | |
| SharedCLIP + TEVIBackbone=ViT-B/162026.06 | 44.3 | 70.97 | 79.53 | |
| SigLIP + TEVIBackbone=ViT-B/162026.06 | 43.77 | 71.56 | 80.73 | |
| SharedCLIPBackbone=ViT-B/162026.06 | 43.31 | 69.53 | 78.86 | |
| CLIPBackbone=ViT-B/162026.06 | 42.46 | 70.33 | 79.31 | |
| AlignCLIP + TEVIBackbone=ViT-B/162026.06 | 42.31 | 70.33 | 78.92 | |
| AlignCLIPBackbone=ViT-B/162026.06 | 41.91 | 70.01 | 78.97 |