Image-to-Text Retrieval on UCM (val)
63.5R@5CLIP (oracle)
Evaluation Results
| Method | Links | |
|---|---|---|
| CLIP (oracle)Data=L + U2026.04 | 63.5 | |
| ToMAData=L=U2026.04 | 61.4 | |
| ToMAData=L≠U2026.04 | 60.6 | |
| ToMA-domainData=L=U2026.04 | 59.2 | |
| ToMA-domainData=L≠U2026.04 | 58.7 | |
| SemiCLIPData=L≠U2026.04 | 57.9 | |
| SemiCLIPData=L=U2026.04 | 57.3 | |
| S-CLIPData=L=U2026.04 | 57 | |
| S-CLIPData=L≠U2026.04 | 55.7 | |
| CLIP (fine-tuned)Data=L2026.04 | 55.2 | |
| Hard-PLData=L=U2026.04 | 54.8 | |
| Soft-PLData=L=U2026.04 | 54.8 | |
| Soft-PLData=L≠U2026.04 | 54.1 | |
| Hard-PLData=L≠U2026.04 | 52.7 | |
| CLIP (original)Data=-2026.04 | 46.7 | |
| S-CLIPData setup=L=U2023.05 | 45.7 | |
| S-CLIPData setup=L!=U2023.05 | 43.5 | |
| Soft-PLData setup=L=U2023.05 | 43.2 | |
| Hard-PLData setup=L!=U2023.05 | 42.1 | |
| CLIP (fine-tune)Data setup=L2023.05 | 41.3 | |
| Hard-PLData setup=L=U2023.05 | 40.8 | |
| Soft-PLData setup=L!=U2023.05 | 40.2 | |
| CLIP (original)Setup=Original2023.05 | 34.3 |