Artistic Image Retrieval on WeArt Query 20235 Values 177338 (test)
70.3mAP@1StyleDecoupler
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| StyleDecouplerBackbone=CLIP2026.01 | 70.3 | 62.3 | 34.3 | 91.6 | 98.8 | 63.9 | 67.6 | |
| StyleDecouplerBackbone=SigLIP2026.01 | 69.9 | 62.2 | 34.6 | 91.4 | 99.1 | 64.5 | 67.3 | |
| CLIPBackbone=ViT-L2026.01 | 66.8 | 54.4 | 30.4 | 88.4 | 96.5 | 52.5 | 59.4 | |
| SigLIP2026.01 | 66.4 | 55.3 | 30.4 | 89.2 | 97.8 | 52.3 | 59.7 | |
| CLIPBackbone=ViT-L, Fine-tuning=true2026.01 | 63.7 | 53.9 | 35.5 | 84.7 | 95.2 | 41.1 | 49.9 | |
| CSDBackbone=ViT-L2026.01 | 60.2 | 49.2 | 33.6 | 82.5 | 91.3 | 37.8 | 44.4 | |
| GDABackbone=ViT-L2026.01 | 52.9 | 44.1 | 27.4 | 75.7 | 89.5 | 39.2 | 41 | |
| ALADINBackbone=ViT2026.01 | 47.5 | 38.2 | 25.7 | 72.9 | 86.3 | 36 | 39.7 | |
| MoCov3Backbone=ViT-L2026.01 | 37.3 | 32.5 | 16.6 | 77.6 | 91.1 | 33.6 | 36.9 | |
| DINOv2Backbone=ViT-L2026.01 | 36.4 | 32.1 | 16.4 | 77 | 91 | 33.1 | 35.4 |