Research map / 3D vision / Depth and stereo
U-ViLAR: Uncertainty-Aware Visual Localization for Autonomous Driving via Differentiable Association and Registration
Xiaofan Li, Zhihao Xu, Chenming Wu, Zhao Yang, Yumeng Zhang, Jiang-Jiang Liu, Haibao Yu, Xiaoqing Ye, Yuan Wang, Shirui Li, Xun Sun, Ji Wan and 1 more
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 U-ViLAR instead of BEV-Locator for HD-map localization | multi-view localization · nuScenes + HD map | orientation MAE (deg): 0.510 -> 0.075 | Improved |
| use U-ViLAR instead of OrienterNet (monocular) | monocular localization · nuScenes + HD map | longitudinal MAE (m): 0.279 -> 0.223 | Improved |
| use U-ViLAR instead of ICP-based method | multi-view localization · SRoad + HD map | longitudinal MAE (m): 0.614 -> 0.284 | Improved |
| use U-ViLAR instead of OrienterNet with OSM | monocular relocalization · KITTI + OSM | orientation recall@1deg (%): 20.41 -> 64.92 | Improved |
| use U-ViLAR instead of MapLocNet (multi-view) | relocalization with OSM · nuScenes + OSM | position recall@1m (%): 20.10 -> 34.53 | Improved |
| add PU-Guided Association (ablation) | U-ViLAR registration-only model · nuScenes + HD map | lateral MAE (m): 0.058 -> 0.040 | Improved |
| add LU-Guided Registration (ablation) | U-ViLAR association-only model · nuScenes + HD map | lateral MAE (m): 0.047 -> 0.040 | Improved |
| add both association and registration (ablation) | U-ViLAR baseline · nuScenes + HD map | orientation MAE (deg): 0.360 -> 0.075 | Improved |
| add Perceptual Uncertainty to the similarity matrix (ablation) | PU-Guided Association · nuScenes + HD map | orientation MAE (deg): 0.094 -> 0.075 | Improved |
| add Local Association (ablation) | PU-Guided Association · nuScenes + HD map | orientation MAE (deg): 0.081 -> 0.075 | Improved |
| add pose distribution prior to registration (ablation) | LU-Guided Registration · nuScenes + HD map | longitudinal RMSE (m): 0.184 -> 0.158 | Improved |
| add pose distribution prior to registration (ablation) | LU-Guided Registration · nuScenes + HD map | lateral MAE (m): 0.042 -> 0.040 | Improved |
Closest papers by meaning
- MapUQ: Map with Uncertainty Quantification for Robust BEV Vectorized Construction (ICML 2026 · similarity 0.85)
- Uncertainty-Aware Vision-Based Metric Cross-View Geolocalization (CVPR 2023 · similarity 0.85)
- V2VLoc: Robust GNSS-Free Collaborative Perception via LiDAR Localization (AAAI 2026 · similarity 0.85)
- Map It Anywhere: Empowering BEV Map Prediction using Large-scale Public Datasets (NeurIPS 2024 · similarity 0.84)
- Uncertainty-Instructed Structure Injection for Generalizable HD Map Construction (CVPR 2025 · similarity 0.83)
- ERMVP: Communication-Efficient and Collaboration-Robust Multi-Vehicle Perception in Challenging Environments (CVPR 2024 · similarity 0.83)
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.
