Text-to-Image Retrieval on Sydney (val)
65.4R@5CLIP (oracle)
Evaluation Results
| Method | Links | |
|---|---|---|
| CLIP (oracle)Data=L + U2026.04 | 65.4 | |
| ToMA-domainData=L=U2026.04 | 62.6 | |
| ToMA-domainData=L≠U2026.04 | 60.9 | |
| SemiCLIPData=L≠U2026.04 | 59.2 | |
| SemiCLIPData=L=U2026.04 | 58.6 | |
| ToMAData=L≠U2026.04 | 56.9 | |
| Soft-PLData=L=U2026.04 | 56.3 | |
| ToMAData=L=U2026.04 | 56.3 | |
| CLIP (fine-tune)Data setup=L2023.05 | 56.1 | |
| Hard-PLData=L=U2026.04 | 55.2 | |
| S-CLIPData setup=L=U2023.05 | 55.1 | |
| Soft-PLData setup=L!=U2023.05 | 54 | |
| S-CLIPData=L≠U2026.04 | 52.9 | |
| S-CLIPData setup=L!=U2023.05 | 52.3 | |
| Soft-PLData=L≠U2026.04 | 52.3 | |
| CLIP (original)Setup=Original2023.05 | 51.7 | |
| Hard-PLData=L≠U2026.04 | 51.7 | |
| S-CLIPData=L=U2026.04 | 50.6 | |
| Soft-PLData setup=L=U2023.05 | 50.2 | |
| Hard-PLData setup=L!=U2023.05 | 50 | |
| CLIP (fine-tuned)Data=L2026.04 | 49.4 | |
| Hard-PLData setup=L=U2023.05 | 47.7 | |
| CLIP (original)Data=-2026.04 | 44.8 |