Object Detection on UOBD v1 (test)
89.3KITTIDetection Hub
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Detection HubBackbone=ResNet-50, Training Protocol=jointly trained on UODB2022.06 | 89.3 | 81.3 | 63.7 | 97 | 94.6 | 65.2 | 62.5 | 49.7 | 63.5 | 57.1 | 71 | 56.8 | — | |
| GAIA-TSASpre-training dataset=GAIA, task-specific architecture selection=true2021.06 | 75.6 | 87.4 | 62.7 | 92.1 | 90.1 | 70.8 | 62.1 | 61.1 | 72.2 | 69.7 | 74.4 | — | — | |
| GAIApre-training dataset=GAIA2021.06 | 72.9 | 85.9 | 62.6 | 91.2 | 89.8 | 69.2 | 59.4 | 57 | 67.9 | 63.5 | 71.9 | — | — | |
| Baselinepre-training dataset=COCO2021.06 | 70.2 | 84.2 | 61.4 | 88.6 | 88.7 | 64.7 | 52.4 | 51.6 | 56.9 | 56.1 | 67.5 | — | — | |
| Single-DINONumber of Parameters=46.5M2026.04 | 69 | 57.8 | 39.3 | 76.9 | 47.3 | 44.1 | 40.5 | 13.9 | 11.7 | 16.3 | 40.8 | 42.6 | 31.3 | |
| DApre-training dataset=UODB original baseline, Domain Adaptation=true2021.06 | 68 | 82.4 | 51.3 | 87.6 | 90 | 56.3 | 53.4 | 53.4 | 55.8 | 60.6 | 65.9 | — | — | |
| UniDABackbone=ResNet-502022.06 | 68 | 82.4 | 51.3 | 87.6 | 90 | 56.3 | 53.4 | 53.4 | 55.8 | 60.6 | 64.2 | 47 | — | |
| Baselinestatus=re-implemented, pre-training dataset=ImageNet2021.06 | 67.1 | 81.5 | 62.1 | 90 | 89.5 | 68.3 | 57.4 | 45.5 | 31.2 | 53.4 | 64.6 | — | — | |
| GEM-DINONumber of Parameters=47.3M2026.04 | 67.1 | 58.1 | 40.4 | 74.2 | 51.4 | 46 | 43.8 | 27.5 | 26.8 | 30.9 | 44.8 | 41.5 | 29.2 | |
| MoE++Number of Parameters=48.1M2026.04 | 66.5 | 57.4 | 39.2 | 73.6 | 50.8 | 45.7 | 43.4 | 25.8 | 25.2 | 28 | 43.6 | 40.5 | 27.9 | |
| GEMGR-DINONumber of Parameters=49.7M2026.04 | 66.5 | 57.9 | 40.1 | 73.4 | 51 | 45.7 | 43 | 26.2 | 26.4 | 31.5 | 44.2 | 40.6 | 28.4 | |
| REMoENumber of Parameters=47.6M2026.04 | 66.2 | 57.8 | 39.5 | 73.4 | 51.7 | 45.6 | 42.9 | 26.3 | 25.7 | 29.2 | 44 | 41 | 28.3 | |
| μMoENumber of Parameters=47.0M2026.04 | 65.9 | 57.3 | 38.7 | 72.9 | 49.5 | 44.8 | 42.5 | 25.7 | 25.2 | 28.6 | 43.3 | 40.3 | 28.5 | |
| Sinkhorn-MoENumber of Parameters=48.3M2026.04 | 65.4 | 57 | 40.2 | 73.1 | 50.2 | 45 | 42.7 | 26.3 | 26 | 29.3 | 43.5 | 39.7 | 27.6 | |
| DAMEXNumber of Parameters=47.5M2026.04 | 65.1 | 56.8 | 39 | 72.6 | 49.4 | 44.9 | 43 | 26.1 | 24.8 | 27.6 | 43.2 | 40.2 | 27.2 | |
| G-DINONumber of Parameters=56.9M2026.04 | 64.6 | 57.9 | 38.5 | 71.8 | 49.7 | 45.8 | 42.4 | 25.9 | 25.1 | 28.3 | 43.4 | 40.6 | 30.2 | |
| Baselinepre-training dataset=UODB original baseline2021.06 | 64.3 | 78.5 | 48.8 | 88.3 | 87.7 | 57.5 | 51.2 | 45.8 | 32.1 | 52.6 | 60.7 | — | — | |
| SingleBackbone=ResNet-50, Training Protocol=trained on each single dataset2022.06 | 64.3 | 78.5 | 48.9 | 88.3 | 87.7 | 57.5 | 51.3 | 45.8 | 32.1 | 52.4 | 59.4 | 47.3 | — | |
| DINO-MoENumber of Parameters=47.5M2026.04 | 64.3 | 57.2 | 38.6 | 72.3 | 50.2 | 44.3 | 42.6 | 25.7 | 24.2 | 26.5 | 42.7 | 39.5 | 27.5 | |
| Mixed-DINONumber of Parameters=46.8M2026.04 | 63.7 | 55.6 | 37.5 | 71.5 | 48.2 | 43.7 | 42.3 | 24.2 | 23.5 | 27 | 41.9 | 39 | 26.9 | |
| Shi et al.Number of Parameters=43.5M2026.04 | 61.5 | 53.7 | 35.4 | 68.2 | 46.5 | 41.5 | 38.3 | 24.8 | 24.2 | 27.5 | 40.6 | 38.3 | 27.2 | |
| Zhou et al.Number of Parameters=69.3M2026.04 | 49.2 | 31.8 | 27.5 | 61.8 | 43.1 | 29.5 | 23.2 | 24.1 | 23.9 | 26.3 | 32.7 | 19.7 | — |