Research map / Label-efficient and robust learning / Recognition and long-tail
Online Map Vectorization for Autonomous Driving: A Rasterization Perspective
Gongjie Zhang, Jiahao Lin, Shuang Wu, Yilin Song, Zhipeng Luo, Yang Xue, Shijian Lu, Zuoguan Wang
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 MapVR rasterized supervision to MapTR | MapTR · nuScenes Map (basic) val, camera, 24 epochs | AP_raster avg.: 24.3 -> 31.2 | Improved |
| Add MapVR rasterized supervision to MapTR | MapTR · nuScenes Map (basic) val, camera, 110 epochs | AP_raster avg.: 35.0 -> 38.5 | Improved |
| Add MapVR rasterized supervision to MapTR | MapTR · nuScenes Map (extended) val | AP_raster avg.: 21.1 -> 27.8 | Improved |
| Add MapVR rasterized supervision to MapTR | MapTR · Argoverse2 val | AP_raster avg.: 26.2 -> 30.1 | Improved |
| Add MapVR rasterized supervision to MapTR | MapTR · 6V-mini-v0.4 | AP_raster lane: 41.3 -> 50.8 | Improved |
| Use coarse 64x32 rasterization instead of the default 256x128 (ablation) | MapTR + MapVR · nuScenes Map (basic) val | mAP_raster: 31.2 -> 21.5 | Hurt |
| Set line rasterization softness τ to 0.5 rather than the default 2.0 (ablation) | MapTR + MapVR · nuScenes Map (basic) val, divider | mAP_raster: 33.1 -> 29.2 | Hurt |
| Add self-based direction regularization on predicted points (ablation) | MapTR + MapVR · nuScenes Map (basic) val | mAP_raster: 29.5 -> 31.2 | Improved |
| Use ground-truth point directions for regularization instead of self-based regularization (ablation) | MapTR + MapVR · nuScenes Map (basic) val | mAP_raster: 31.2 -> 29.3 | Hurt |
| Use geometry-aware line and polygon rasterization rather than rendering every element as a line (ablation) | MapTR + MapVR · nuScenes Map (basic) val, pedestrian crossing | mAP_raster: 21.8 -> 37.5 | Improved |
| Apply rasterized supervision directly to vector predictions rather than using a parallel segmentation branch (ablation) | MapTR · nuScenes Map (basic) val | mAP_raster: 26.7 -> 31.2 | Improved |
Closest papers by meaning
- VectorMapNet: End-to-end Vectorized HD Map Learning (ICML 2023 · similarity 0.93)
- PriorDrive: Enhancing Online HD Mapping with Unified Vector Priors (AAAI 2026 · similarity 0.90)
- Unveiling the Hidden: Online Vectorized HD Map Construction with Clip-Level Token Interaction and Propagation (NeurIPS 2024 · similarity 0.90)
- MapExpert: Online HD Map Construction with Simple and Efficient Sparse Map Element Expert (AAAI 2025 · similarity 0.89)
- Driving by the Rules: A Benchmark for Integrating Traffic Sign Regulations into Vectorized HD Map (CVPR 2025 · similarity 0.88)
- InteractionMap: Improving Online Vectorized HDMap Construction with Interaction (CVPR 2025 · similarity 0.88)
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.
