Research map / 3D vision / Depth and stereo
Visual Cross-View Metric Localization with Dense Uncertainty Estimates
Zimin Xia, Olaf Booij, Marco Manfredi, Julian F. P. Kooij
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add infoNCE bottleneck loss (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 19.25 -> 14.30 | Improved |
| add Gaussian label smoothing (sigma 4 pixels) (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 14.30 -> 13.39 | Improved |
| increase satellite descriptor grid from N=1 to N=8 (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 19.62 -> 13.39 | Improved |
| increase satellite descriptor grid from N=8 to N=16 (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 13.39 -> 15.04 | Hurt |
| add matching score map M to satellite descriptors (S+G to S+M+G) (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 18.35 -> 13.83 | Improved |
| add ground descriptor G to S+M (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 13.39 -> 13.83 | No clear effect |
| add satellite context to the matching score map (ablation) | Cross-view metric localization model · VIGOR tuning split | Mean error (m): 24.37 -> 13.39 | Improved |
| replace regression with dense probabilistic output | CVR baseline · VIGOR same-area, positives | Median error (m): 9.31 -> 4.58 | Improved |
| replace regression with dense probabilistic output | CVR baseline · VIGOR cross-area, positives | Mean error (m): 11.26 -> 13.06 | Hurt |
| replace regression with dense probabilistic output | CVR baseline · VIGOR cross-area, positives | Median error (m): 10.02 -> 6.31 | Improved |
| replace regression with dense probabilistic output | CVR baseline · Oxford RobotCar | Median error (m): 1.72 -> 1.24 | Improved |
| add CVR-like regression layer on top of dense output (Variation 1) (ablation) | Cross-view metric localization model · VIGOR same-area, positives | Median error (m): 4.58 -> 6.29 | Hurt |
Closest papers by meaning
- View Consistent Purification for Accurate Cross-View Localization (ICCV 2023 · similarity 0.91)
- Beyond Cross-View Image Retrieval: Highly Accurate Vehicle Localization Using Satellite Image (CVPR 2022 · similarity 0.89)
- Cross-Descriptor Visual Localization and Mapping (ICCV 2021 · similarity 0.89)
- Boosting 3-DoF Ground-to-Satellite Camera Localization Accuracy via Geometry-Guided Cross-View Transformer (ICCV 2023 · similarity 0.89)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.87)
- CroCoDL: Cross-device Collaborative Dataset for Localization (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
