Cross-view ground-camera localization on KITTI 2 (test)
11.69Pos. Error (m) MeanCross3R
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Cross3RModel Type=Multi-view 3D reconstruction models, Zero-shot=true, Orientation=unknown, Inputs=ground and satellite images only2026.05 | 11.69 | 10.52 | 24.27 | 3.51 | 39.11 | 76.32 | 85.95 | 3.5 | 10.7 | 19.07 | 16.14 | 44.9 | |
| CCVPEModel Type=2D-only baselines, Zero-shot=true, Orientation=unknown, Inputs=ground and satellite images only2026.05 | 13.94 | 10.98 | 77.84 | 63.84 | 23.42 | 49.15 | 60.46 | 11.81 | 29.85 | 42.12 | 4.39 | 14.09 | |
| SliceMatchModel Type=2D-only baselines, Zero-shot=true, Orientation=unknown, Inputs=ground and satellite images only2026.05 | 14.85 | 11.85 | 23.64 | 7.96 | 24 | 62.52 | 72.89 | 7.17 | 26.11 | 33.12 | 31.69 | 31.69 | |
| LMModel Type=2D-only baselines, Zero-shot=true, Orientation=unknown, Inputs=ground and satellite images only2026.05 | 15.5 | 16.02 | 89.84 | 89.85 | 5.6 | 16.02 | 25.6 | 5.64 | 15.86 | 25.76 | 0.6 | 1.6 | |
| π3*Model Type=Multi-view 3D reconstruction models, Zero-shot=true, Orientation=unknown, Inputs=ground and satellite images only, Training=fine-tuned on CrossGeo2026.05 | 16.28 | 14.85 | 23.16 | 4.09 | 9.44 | 27.61 | 44.18 | 4.34 | 12.86 | 21.36 | 15.27 | 40.27 | |
| π3Model Type=Multi-view 3D reconstruction models, Zero-shot=true, Orientation=unknown, Inputs=ground and satellite images only2026.05 | 48.26 | 42.81 | 27.67 | 13.45 | 4.18 | 10.69 | 18.1 | 2.85 | 7.45 | 11.92 | 5.56 | 16.31 |