Research map / Detection and segmentation / CNN features
ReVMap: Vectorized Global Mapping via Connectivity-Aware Local Map Fusion
Hongjae Shin, Jungho Kim, Donghyuk Kwak, Myeongjun Kim, Seunghoon Yu, Jiyong Oh, Jun Won Choi
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes ReVMap, a local-to-global auto-labeling method for vectorized HD map construction that fuses local map predictions with connectivity awareness to limit error propagation.
Problem
Annotating vectorized HD maps is labor-intensive; incremental local-to-global auto-labeling accumulates inaccurate predictions that are reused as priors, causing persistent error propagation across iterations.
Approach
Connectivity-aware fusion of local vector map predictions into a global map, intended to manage reliability so that inaccurate predictions are not accumulated and reused as priors (based on the abstract only; details unavailable).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add learned connectivity estimation instead of endpoint-distance rule (ablation) | ReVMap baseline · nuScenes new val (onboard C, ablation) | mGAP: 19.4 -> 22.9 | Improved |
| add learned connectivity estimation (ablation) | ReVMap baseline · nuScenes new val (onboard C, ablation) | mAP: 39.0 -> 39.0 | No clear effect |
| add global map prior (ablation) | ReVMap w/ connectivity · nuScenes new val (onboard C, ablation) | mAP: 39.0 -> 44.8 | Improved |
| add point-wise uncertainty estimation (ablation) | ReVMap w/ prior · nuScenes new val (onboard C, ablation) | mGAP: 26.8 -> 27.4 | Improved |
| add uncertainty-aware map update (ablation) | ReVMap w/ uncertainty estimation · nuScenes new val (onboard C, ablation) | mGAP: 27.4 -> 28.5 | Improved |
| add uncertainty filtering only (ablation) | Global map update · nuScenes new val (onboard C, ablation) | mGAP: 27.4 -> 28.1 | Improved |
| add uncertainty-weighted interpolation only (ablation) | Global map update · nuScenes new val (onboard C, ablation) | mGAP: 27.4 -> 27.9 | Improved |
| use perturbed GT boxes for dynamic removal (ablation) | ReVMap C+L · nuScenes new val (offboard C+L) | mGAP: 36.1 -> 35.6 | Hurt |
| skip dynamic-object removal (ablation) | ReVMap C+L · nuScenes new val (offboard C+L) | mGAP: 36.1 -> 35.2 | Hurt |
| swap ReVMap global construction for GlobalMapNet (ablation) | ReVMap local predictions · nuScenes new val | mGAP: 28.5 -> 23.5 | Hurt |
| swap ReVMap global construction for VMA (ablation) | ReVMap local predictions · nuScenes new val | mGAP: 28.5 -> 25.7 | Hurt |
| plug ReVMap global construction into MapTracker local predictions (ablation) | MapTracker · nuScenes new val | mGAP: 24.5 -> 26.5 | Improved |
Closest papers by meaning
- ArgoTweak: Towards Self-Updating HD Maps through Structured Priors (ICCV 2025 · similarity 0.86)
- PriorMaskMap: Robust Online Vectorized Map Construction with Biased Priors (ECCV 2026 · similarity 0.86)
- InteractionMap: Improving Online Vectorized HDMap Construction with Interaction (CVPR 2025 · similarity 0.85)
- Uncertainty-Instructed Structure Injection for Generalizable HD Map Construction (CVPR 2025 · similarity 0.84)
- SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction (NeurIPS 2025 · similarity 0.83)
- Change-Robust Online Topological Memory for Long-Term Relocalization and Semantic Navigation (NeurIPS 2026 · 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.
