Zero-shot Image-Text Retrieval on MSCOCO
82.8AccuracySigLIP2-ViT-B/16
Evaluation Results
| Method | Links | |
|---|---|---|
| SigLIP2-ViT-B/16Backbone=ViT-B/162026.03 | 82.8 | |
| C^2LIPBackbone=ViT-B/162026.03 | 82.7 | |
| SigLIP ViT-L/16-res256Backbone=ViT-L/16, Input Resolution=256x2562026.03 | 82.3 | |
| CLIPS (ViT-L/14)Backbone=ViT-L/142026.03 | 82.1 | |
| BLIP-BBackbone=ViT-B2026.03 | 79.1 | |
| CLIP-A (ViT-L/14)Backbone=ViT-L/142026.03 | 78.1 | |
| FLAVABackbone=ViT-B2026.03 | 73.9 |