Fashion Image Retrieval on LOOKBENCH (nDCG@5)
66.27nDCG@5 (AiGen vs StreetLook)MARQO-FASHIONSIGLIP
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MARQO-FASHIONSIGLIPResolution=224, Embedding Size=7682026.01 | 66.27 | 58.53 | 42.43 | 51.86 | 49.44 | |
| GR-PROResolution=336, Embedding Size=10242026.01 | 63.67 | 54.88 | 44.75 | 51.55 | 49.8 | |
| MARQO-FASHIONCLIPResolution=224, Embedding Size=5122026.01 | 63.22 | 54.93 | 41.87 | 51.68 | 48.63 | |
| GR-LITEResolution=336, Embedding Size=10242026.01 | 62.47 | 52.08 | 43.84 | 51.7 | 49.18 | |
| SIGLIP2-B/16Resolution=384, Embedding Size=7682026.01 | 57.83 | 54.97 | 39.35 | 49.12 | 46.1 | |
| SIGLIP2-L/16Resolution=384, Embedding Size=10242026.01 | 51.89 | 48.57 | 35.91 | 44.78 | 41.86 | |
| PP-SHITUV2Resolution=224, Embedding Size=5122026.01 | 30.06 | 33.69 | 32.77 | 43.22 | 37.17 | |
| CLIP-L/14Resolution=336, Embedding Size=7682026.01 | 25.28 | 25.95 | 21.09 | 40.35 | 30.08 | |
| DINOV2-VIT-LResolution=224, Embedding Size=10242026.01 | 24.29 | 25.05 | 22.99 | 37.66 | 29.57 | |
| DINOV2-VIT-GResolution=224, Embedding Size=15362026.01 | 23.13 | 23.27 | 22.96 | 37.71 | 29.36 | |
| DINOV3-VIT-LResolution=224, Embedding Size=10242026.01 | 20.24 | 27.66 | 26.27 | 39.85 | 31.83 | |
| DINOV3-VIT-7BResolution=224, Embedding Size=40962026.01 | 19.42 | 30.3 | 26.77 | 39.38 | 31.99 | |
| CLIP-B/16Resolution=224, Embedding Size=5122026.01 | 17.86 | 13.75 | 16.8 | 34.75 | 24.36 | |
| INTERNVIT-6BResolution=448, Embedding Size=10242026.01 | 14.17 | 20.13 | 16.4 | 31.33 | 22.99 | |
| DINOV3-CONVNEXTResolution=224, Embedding Size=15362026.01 | 10.45 | 19.13 | 19.12 | 33.38 | 24.68 |