Aerial Visual Localization on UAVD4L LoD (in-Traj.)
97.6Accuracy (2m-2°)LoD-Loc v3
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LoD-Loc v3Method Variation=Full2026.03 | 97.6 | 98.9 | 99.7 | 0.49 | 0.13 | |
| LoD-Loc v3Method Variation=no select2026.03 | 97.4 | 99.1 | 99.8 | 0.5 | 0.12 | |
| LoD-Loc v2Method Variation=Full, In-distribution training=true2026.03 | 93.7 | 98.4 | 99.5 | 0.72 | 0.15 | |
| LoD-Loc v2Method Variation=no select, In-distribution training=true2026.03 | 93.5 | 98.4 | 99.5 | 0.74 | 0.17 | |
| RoMAReference Model=UAVD4L Mesh model2024.10 | 93.27 | 93.7 | 93.77 | — | — | |
| SPP+SPGReference Model=UAVD4L Mesh model2024.10 | 91.71 | 92.02 | 92.14 | — | — | |
| LoFTRReference Model=UAVD4L Mesh model2024.10 | 84.98 | 88.09 | 88.9 | — | — | |
| e-LoFTRReference Model=UAVD4L Mesh model2024.10 | 84.47 | 88.21 | 88.96 | — | — | |
| LoD-LocVariation=Full model2024.10 | 84.41 | 91.77 | 96.95 | — | — | |
| LoD-LocModel Variant=LoD-Loc2026.06 | 83.42 | 90.65 | 96.57 | 1.73 | 0.19 | |
| SIFT+NNReference Model=UAVD4L Mesh model2024.10 | 73.13 | 78.62 | 80.42 | — | — | |
| SemCityLocModel Variant=Full model2026.06 | 70.45 | 92.96 | 99.19 | 1.68 | 0.19 | |
| LoD-LocVariation=no USR2024.10 | 70.39 | 85.47 | 95.32 | — | — | |
| SemCityLocModel Variant=No Coarse Selection2026.06 | 62.03 | 89.96 | 99.05 | 1.84 | 0.2 | |
| LoD-LocVariation=no Refine2024.10 | 51.31 | 76.06 | 86.78 | — | — | |
| LoD-LocIn-distribution training=true2026.03 | 49.56 | 71.82 | 89.09 | 3.32 | 1.48 | |
| SemCityLocModel Variant=No Refinement2026.06 | 47.07 | 74.88 | 91.21 | 2.48 | 0.34 | |
| LoD-LocVariation=no wireframe estimation2024.10 | 10.41 | 16.21 | 24.19 | — | — | |
| RoMAReference Model=CadLoc LoD model2024.10 | 2.18 | 2.87 | 3.68 | — | — | |
| DINOv2Framework=MC-Loc2026.03 | 1.2 | 4.1 | 17.4 | 8.29 | 2.58 | |
| e-LoFTRReference Model=CadLoc LoD model2024.10 | 0.37 | 0.87 | 1.31 | — | — | |
| RoMaFramework=MC-Loc2026.03 | 0.1 | 0.6 | 3.3 | 10.6 | 8.6 | |
| Sensor PriorsModel Type=Sensor Priors2024.10 | 0 | 0 | 4.3 | — | — | |
| SIFT+NNReference Model=CadLoc LoD model2024.10 | 0 | 0 | 0 | — | — | |
| SPP+SPGReference Model=CadLoc LoD model2024.10 | 0 | 0 | 0 | — | — | |
| LoFTRReference Model=CadLoc LoD model2024.10 | 0 | 0 | 0 | — | — | |
| SOLD2Reference Model=CadLoc LoD model2024.10 | 0 | 0 | 0 | — | — | |
| DeepLSD+SOLD2Reference Model=CadLoc LoD model2024.10 | 0 | 0 | 0 | — | — | |
| DeepLSD+GlueStickReference Model=CadLoc LoD model2024.10 | 0 | 0 | 0 | — | — | |
| Prior2026.03 | 0 | 0 | 4.3 | 6.48 | 1.63 | |
| SIFT+NNFramework=CAD-Loc2026.03 | 0 | 0 | 0 | — | — | |
| SPP+SPGFramework=CAD-Loc2026.03 | 0 | 0 | 0 | — | — | |
| LoFTRFramework=CAD-Loc2026.03 | 0 | 0 | 0 | — | — | |
| e-LoFTRFramework=CAD-Loc2026.03 | 0 | 0 | 0 | — | — | |
| RoMAFramework=CAD-Loc2026.03 | 0 | 0 | 0 | — | — | |
| LoD-Loc v2Method Variation=no refine, In-distribution training=true2026.03 | 0 | 0 | 23.38 | 6.19 | 0.67 | |
| LoD-Loc v3Method Variation=no refine2026.03 | 0 | 0 | 24.1 | 6.09 | 0.67 |