Research map / Detection and segmentation / CNN features
Online Vectorized HD Map Construction using Geometry
Zhixin Zhang, Yiyuan Zhang, Xiaohan Ding, Fusheng Jin, Xiangyu Yue
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 geometry-decoupled attention (GDA) to the baseline decoder (ablation) | GeMap baseline · nuScenes (24 epochs) | mAP: 49.3 -> 51.2 | Improved |
| add Euclidean shape and relation losses without GDA (ablation) | GeMap baseline · nuScenes (24 epochs) | mAP: 49.3 -> 48.8 | Hurt |
| add only relation loss or only shape loss without GDA (ablation) | GeMap baseline · nuScenes (24 epochs) | mAP: 49.3 -> 49.7 / 49.4 | No clear effect |
| add Euclidean losses on top of GDA (ablation) | GeMap with GDA · nuScenes (24 epochs) | mAP: 51.2 -> 52.6 | Improved |
| use GDA instead of 2 layers of vanilla self-attention (ablation) | GeMap decoder · nuScenes (24 epochs) | mAP: 51.4 -> 52.6 | Improved |
| use both shape and relation losses vs a single loss with GDA (ablation) | GeMap with GDA · nuScenes (24 epochs) | mAP: 52.4 -> 52.6 | Mixed |
| GeMap vs MapTRv2 on Argoverse 2 (R50, dense loss) | online HD map construction · Argoverse 2 | mAP: 67.4 -> 71.8 | Improved |
| GeMap Swin-T vs MapTRv2 V2-99 backbone | online HD map construction · nuScenes | mAP: 73.4 -> 72.0 | Hurt |
| GeMap vs MapTRv2 FPS (R50, dense loss) | online HD map construction · nuScenes | FPS: 15.0 -> 15.8 | Improved |
Closest papers by meaning
- Unveiling the Hidden: Online Vectorized HD Map Construction with Clip-Level Token Interaction and Propagation (NeurIPS 2024 · similarity 0.89)
- HIMap: HybrId Representation Learning for End-to-end Vectorized HD Map Construction (CVPR 2024 · similarity 0.88)
- InteractionMap: Improving Online Vectorized HDMap Construction with Interaction (CVPR 2025 · similarity 0.88)
- MGMap: Mask-Guided Learning for Online Vectorized HD Map Construction (CVPR 2024 · similarity 0.87)
- MapExpert: Online HD Map Construction with Simple and Efficient Sparse Map Element Expert (AAAI 2025 · similarity 0.86)
- MGMapNet: Multi-Granularity Representation Learning for End-to-End Vectorized HD Map Construction (ICLR 2025 · similarity 0.86)
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.
