Research map / 3D vision / Depth and stereo
VectorReLoc: Reliable Vectorized SD Map Visual Re-localization with Contrastive Feature Alignment
Ziming Liu, Quanjie Xiang, Wang yun, wang yiting, chao wen, Zhuanjian XU, Leichen Wang, HAO SUN, Guangyu Gao
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 pretraining on the VectorReLoc dataset before fine-tuning (ablation) | VectorReLoc · nuScenes | position recall @1m (%): 39.09 -> 65.79 | Improved |
| add pretraining on the VectorReLoc dataset before fine-tuning (ablation) | VectorReLoc · Argoverse2 | position recall @1m (%): 49.86 -> 58.05 | Improved |
| add pretraining on the VectorReLoc dataset before fine-tuning (ablation) | VectorReLoc · nuScenes | Epos (m): 2.15 -> 1.36 | Improved |
| add contrastive joint training to regression-only (ablation) | VectorReLoc · VectorReLoc dataset (real GNSS noise) | position recall @1m (%): 22.96 -> 57.09 | Improved |
| add contrastive joint training to regression-only (ablation) | VectorReLoc · VectorReLoc dataset (real GNSS noise) | xy error (m): 3.0632 -> 1.3740 | Improved |
| add contrastive joint training to regression-only (ablation) | VectorReLoc · VectorReLoc dataset (simulated GNSS noise) | position recall @1m (%): 5.25 -> 19.27 | Improved |
| train contrastive and regression objectives jointly instead of pretrain-then-finetune (ablation) | VectorReLoc · VectorReLoc dataset (real GNSS noise) | position recall @1m (%): 52.40 -> 57.09 | Improved |
| add reliability filtering at threshold 0.9 (ablation) | VectorReLoc · VectorReLoc dataset | yaw error (deg): 0.3620 -> 0.2712 | Improved |
| train on synthetic Student-t offsets instead of real GNSS offsets (ablation) | VectorReLoc · VectorReLoc dataset test split | Epos and 5 m recall: <=0.01 m in Epos and <=0.07 pp in 5 m recall | No clear effect |
| increase simulated training offset range from 1 m to 30 m (ablation) | VectorReLoc · Argoverse2 | Epos (m): 1.04 -> 1.74 (1 m to 13 m); 1.67 at 30 m | Hurt |
| quantize to Int8 instead of Float32 | VectorReLoc · Horizon J6M deployment | position recall @1m (%): 26.25 -> 17.35 | Hurt |
| quantize to Int16 instead of Float32 | VectorReLoc · Horizon J6M deployment | position recall @1m (%): 26.25 -> 26.05 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OrienterNet — OrienterNet: Visual Localization in 2D Public Maps With Neural Matching (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- V2VLoc: Robust GNSS-Free Collaborative Perception via LiDAR Localization (AAAI 2026 · similarity 0.86)
- AutoCompass: Accurate Visual Localization on Public Maps by Learning from Weak Labels (ECCV 2026 · similarity 0.86)
- Beyond Cross-View Image Retrieval: Highly Accurate Vehicle Localization Using Satellite Image (CVPR 2022 · similarity 0.86)
- SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction (NeurIPS 2025 · similarity 0.86)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 2026 · similarity 0.85)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.84)
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.
