Multi-conditional Image Retrieval on CLAY
44.7mAP (Color Category)CLAY
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CLAYBackbone=SigLIP-B2026.04 | 44.7 | 55 | 49 | 55.6 | 81.5 | 52 | |
| CLAYBackbone=CLIP-B2026.04 | 35.9 | 38.2 | 57 | 56.6 | 78.6 | 58 | |
| SigLIP-BBackbone=SigLIP-B2026.04 | 13.9 | 19.7 | 34.5 | 32.1 | 50.5 | 39.5 | |
| InstructBLIP2026.04 | 12.5 | 14.7 | 38.6 | 31.2 | 74.7 | 44.1 | |
| CLIP-BBackbone=CLIP-B2026.04 | 11.6 | 10.9 | 35.9 | 27.1 | 37 | 31.9 |