Image-to-Text Retrieval on Flickr (T@1, T@10)
19.2T@1ItemizedCLIP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ItemizedCLIPEvaluation Protocol=Zero-shot, Training Data=Itemized-cc0.3M2025.12 | 19.2 | 51.8 | |
| FLAIREvaluation Protocol=Zero-shot, Training Data=Itemized-cc0.3M2025.12 | 16.7 | 46 | |
| DreamLIPEvaluation Protocol=Zero-shot, Training Data=Itemized-cc0.3M2025.12 | 13.6 | 39.2 | |
| Multi-positive SigLIPEvaluation Protocol=Zero-shot, Training Data=Itemized-cc0.3M2025.12 | 11.9 | 36.9 | |
| Vanilla SigLIPEvaluation Protocol=Zero-shot, Training Data=Itemized-cc0.3M2025.12 | 6.9 | 25.7 | |
| Vanilla CLIPEvaluation Protocol=Zero-shot, Training Data=Itemized-cc0.3M2025.12 | 4.7 | 18.4 |