Zero-shot Image-Text Retrieval on Flickr30k
97.7Accuracy (Zero-shot)CLIPS (ViT-L/14)
Evaluation Results
| Method | Links | |
|---|---|---|
| CLIPS (ViT-L/14)Backbone=ViT-L/142026.03 | 97.7 | |
| BLIP-BBackbone=ViT-B2026.03 | 97.2 | |
| C^2LIPBackbone=ViT-B/162026.03 | 97 | |
| SigLIP ViT-L/16-res256Backbone=ViT-L/16, Input Resolution=256x2562026.03 | 96.7 | |
| SigLIP2-ViT-B/16Backbone=ViT-B/162026.03 | 96.5 | |
| CLIP-A (ViT-L/14)Backbone=ViT-L/142026.03 | 95.2 | |
| FLAVABackbone=ViT-B2026.03 | 91.8 |