Research map / Detection and segmentation / CNN features
InsMapper: Exploring Inner-instance Information for Vectorized HD Mapping
zhenhua xu, Kwan-Yee K. Wong, Hengshuang Zhao
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 inner-instance self-attention aggregation after cross-attention (ablation) | InsMapper (MapTR-based) · NuScenes val | mAP: 45.42 -> 48.31 | Improved |
| use restricted inner-instance attention instead of vanilla self-attention (ablation) | InsMapper · NuScenes val | mAP: 46.23 -> 48.31 | Improved |
| add random blocking in inner-instance attention (ablation) | InsMapper · NuScenes val | mAP: 47.67 -> 48.31 | Improved |
| place aggregation module after cross-attention rather than before (ablation) | InsMapper · NuScenes val | mAP: 45.24 -> 48.31 | Improved |
| use self-attention for inner-instance query fusion instead of mean or FFN fusion (ablation) | InsMapper · NuScenes val | mAP | Improved |
| use hybrid query generation instead of basic or hierarchical schemes (ablation) | InsMapper · NuScenes val | mAP | Improved |
| apply InsMapper modules on MapTR | MapTR (24 epochs, R50) · NuScenes val | mAP: 42.93 -> 48.31 | Improved |
| apply InsMapper modules on MapTR-V2 | MapTR-V2 (24 epochs, R50) · NuScenes val | mAP: 51.37 -> 53.72 | Improved |
| InsMapper (MapTR-based) versus MapTR-V2 at 24 epochs | InsMapper · NuScenes val | mAP: 51.37 vs 48.31 | Hurt |
| add inner-instance modules to MapTR (inference speed) | MapTR · inference FPS | FPS: 8.5 -> 7.5 | Hurt |
Closest papers by meaning
- InteractionMap: Improving Online Vectorized HDMap Construction with Interaction (CVPR 2025 · similarity 0.89)
- Unveiling the Hidden: Online Vectorized HD Map Construction with Clip-Level Token Interaction and Propagation (NeurIPS 2024 · similarity 0.87)
- Uncertainty-Instructed Structure Injection for Generalizable HD Map Construction (CVPR 2025 · similarity 0.87)
- MapExpert: Online HD Map Construction with Simple and Efficient Sparse Map Element Expert (AAAI 2025 · similarity 0.86)
- SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction (NeurIPS 2025 · similarity 0.85)
- Compact HD Map Construction via Douglas-Peucker Point Transformer (AAAI 2024 · similarity 0.85)
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.
