Image Retrieval on GLD v2 (Public/Private Evaluation)
58.6mAP@100AMES
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AMESLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 58.6 | — | — | |
| RRTLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 57.1 | — | — | |
| ELViSLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 56.9 | — | — | |
| R2FormerLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 56 | — | — | |
| No re-rankingLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 49.6 | — | — | |
| ChamferLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 42.3 | — | — | |
| Chamfer+OTLocal descriptor backbone=DINOv3, Training domain=GLDv22026.03 | 41.7 | — | — | |
| AMESTraining Domain=landmarks domain (GLDv2), Backbone=DINOv2, Descriptor Category=descriptor-based2026.03 | 34.7 | — | — | |
| RRTTraining Domain=landmarks domain (GLDv2), Backbone=DINOv2, Descriptor Category=descriptor-based2026.03 | 33.1 | — | — | |
| R2FormerTraining Domain=landmarks domain (GLDv2), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 32.6 | — | — | |
| ELViSTraining Domain=landmarks domain (GLDv2), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 32.2 | — | — | |
| No re-rankingTraining Domain=landmarks domain (GLDv2), Backbone=DINOv22026.03 | 27.3 | — | — | |
| No re-rankingTraining Domain=household items domain (SOP), Backbone=DINOv22026.03 | 27.3 | — | — | |
| ChamferTraining Domain=landmarks domain (GLDv2), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 23.8 | — | — | |
| Chamfer+OTTraining Domain=landmarks domain (GLDv2), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 23.8 | — | — | |
| R2FormerTraining Domain=household items domain (SOP), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 23.7 | — | — | |
| ELViSTraining Domain=household items domain (SOP), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 22.9 | — | — | |
| Chamfer+OTTraining Domain=household items domain (SOP), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 18.6 | — | — | |
| AMESTraining Domain=household items domain (SOP), Backbone=DINOv2, Descriptor Category=descriptor-based2026.03 | 17.2 | — | — | |
| ChamferTraining Domain=household items domain (SOP), Backbone=DINOv2, Descriptor Category=similarity-based2026.03 | 15.4 | — | — | |
| RRTTraining Domain=household items domain (SOP), Backbone=DINOv2, Descriptor Category=descriptor-based2026.03 | 10.8 | — | — | |
| CLIPBackbone=ViT-B/322023.04 | — | 7.5 | 7.5 | |
| CLIPBackbone=ViT-B/162023.04 | — | 8.4 | 10.6 | |
| CLIPBackbone=ViT-L/142023.04 | — | 12.1 | 13.6 | |
| CLIPBackbone=ViT-L/14, Resolution=336x3362023.04 | — | 17 | 15.6 | |
| OpenCLIPBackbone=ViT-B/322023.04 | — | 9.17 | 8.4 | |
| OpenCLIPBackbone=ViT-B/162023.04 | — | 12.1 | 11 | |
| OpenCLIPBackbone=ViT-L/142023.04 | — | 14.5 | 13.8 | |
| OpenCLIPBackbone=ViT-H/142023.04 | — | 17.7 | 15.3 | |
| UNICOMBackbone=ViT-B/322023.04 | — | 15.1 | 13.3 | |
| UNICOMBackbone=ViT-B/162023.04 | — | 18.7 | 17.2 | |
| UNICOMBackbone=ViT-L/142023.04 | — | 21.1 | 20.1 | |
| UNICOMBackbone=ViT-L/14, Resolution=336x3362023.04 | — | 23.2 | 22 |