Research map / 3D vision / Depth and stereo
View Consistent Purification for Accurate Cross-View Localization
Shan Wang, Yanhao Zhang, Akhil Perincherry, Ankit Vora, Hongdong Li
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 |
|---|---|---|---|
| use two confidence maps (view-consistent and on-ground) instead of one merged map (ablation) | PureACL · FordAV-CVL Log4 | yaw mean error (deg): 2.08 -> 0.90 | Improved |
| add camera-aware spatial embedding (ablation) | PureACL · FordAV-CVL Log4 | longitudinal mean error (m): 1.17 -> 0.88 | Improved |
| use view-consistent on-ground keypoint detector instead of SuperPoint (ablation) | PureACL · FordAV-CVL Log4 | yaw mean error (deg): 1.05 -> 0.74 | Improved |
| use highest-confidence multi-camera fusion instead of mean fusion (ablation) | PureACL · FordAV-CVL Log4 | yaw mean error (deg): 0.90 -> 0.74 | Improved |
| increase onboard cameras from 1 front to 4 (ablation) | PureACL · FordAV-CVL Log4 | yaw mean error (deg): 2.77 -> 0.74 | Improved |
| use two side cameras instead of front and rear cameras (ablation) | PureACL · FordAV-CVL Log4 | yaw mean error (deg): 1.14 -> 3.56 | Hurt |
| sparse visual-only method versus LiDAR-visual SIBCL on translation | PureACL · KITTI-CVL | lateral mean error (m): 1.02 -> 0.14 | Improved |
| sparse visual-only method versus LiDAR-visual SIBCL on orientation | PureACL · KITTI-CVL | yaw mean error (deg): 1.91 -> 3.57 | Hurt |
| PureACL generalization to unseen road | PureACL · FordAV-CVL Log4->5 | longitudinal mean error (m): 6.27 -> 1.43 | Improved |
| larger initial pose perturbation up to 60 degree yaw or 15 m shift (ablation) | PureACL · KITTI-CVL initial pose | lateral mean error (m): 0.17 -> 3.98 (15m, 15 deg versus 15m, 7.5 deg) | Hurt |
Closest papers by meaning
- Visual Cross-View Metric Localization with Dense Uncertainty Estimates (ECCV 2022 · similarity 0.91)
- Beyond Cross-View Image Retrieval: Highly Accurate Vehicle Localization Using Satellite Image (CVPR 2022 · similarity 0.89)
- View From Above: Orthogonal-View aware Cross-view Localization (CVPR 2024 · similarity 0.88)
- Enhancing Visual Localization with Cross-Domain Image Generation (ICML 2025 · similarity 0.87)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.87)
- BevSplat: Resolving Height Ambiguity via Feature-Based Gaussian Primitives for Weakly-Supervised Cross-View Localization (NeurIPS 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.
