Research map / Vision-language and multimodal / Cross-modal learning
UniMapGen: A Generative Framework for Large-Scale Map Construction from Multi-modal Data
Yujian Yuan, Changjie Wu, Xinyuan Chang, Sijin Wang, Hang Zhang, Shiyi Liang, Shuang Zeng, Mu Xu
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 state update (ablation) | UniMapGen · OpenSatMap20 val | mIoU: 38.99 -> 39.37 | Improved |
| add state update on top of reordering (ablation) | UniMapGen · OpenSatMap20 val | mIoU: 40.08 -> 41.81 | Improved |
| add data augmentation (ablation) | UniMapGen · OpenSatMap20 val | mIoU: 36.84 -> 38.99 | Improved |
| add line reordering (ablation) | UniMapGen · OpenSatMap20 val | mIoU: 39.37 -> 41.81 | Improved |
| add line reordering without state update (ablation) | UniMapGen · OpenSatMap20 val | AP^C_0.9: 24.53 -> 27.96 | Improved |
| add text prompt to BEV for target map (ablation) | UniMapGen · OpenSatMap20 val subset, target GT | mIoU: 16.89 -> 35.56 | Improved |
| add PV frames to BEV+text for target map (ablation) | UniMapGen · OpenSatMap20 val subset, target GT | mIoU: 35.56 -> 41.02 | Improved |
| add PV frames to BEV for full map (ablation) | UniMapGen · OpenSatMap20 val subset, full GT | mIoU: 43.08 -> 45.16 | Improved |
| add PV frames to BEV for full map (ablation) | UniMapGen · OpenSatMap20 val subset, full GT | AP^C_0.9: 29.80 -> 31.91 | Improved |
| generative sequence model vs SegNeXt segmentation | UniMapGen vs SegNeXt · OpenSatMap20 val | mIoU: 33.69 -> 41.81 | Improved |
| generative model vs SegNeXt on strict mask AP75 | UniMapGen vs SegNeXt · OpenSatMap20 val | AP^M_75: 5.26 vs 5.19 | No clear effect |
| generative model vs LaneGAP on landmark F1 | UniMapGen vs LaneGAP · nuScenes lane topology | Landmark F1: 53.2 -> 50.9 | Hurt |
Closest papers by meaning
- OpenSatMap: A Fine-grained High-resolution Satellite Dataset for Large-scale Map Construction (NeurIPS 2024 · similarity 0.90)
- Unveiling the Hidden: Online Vectorized HD Map Construction with Clip-Level Token Interaction and Propagation (NeurIPS 2024 · similarity 0.87)
- SDTagNet: Leveraging Text-Annotated Navigation Maps for Online HD Map Construction (NeurIPS 2025 · similarity 0.87)
- Patched Line Segment Learning for Vector Road Mapping (AAAI 2024 · similarity 0.86)
- Uncertainty-Instructed Structure Injection for Generalizable HD Map Construction (CVPR 2025 · similarity 0.85)
- MAPLM: A Real-World Large-Scale Vision-Language Benchmark for Map and Traffic Scene Understanding (CVPR 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.
