Kuang Wu · Sulei Nian · Can Shen · Chuan Yang · Zhanbin Li
Langge Technology
Feb. 27th, 2025: 👏 Check out our latest work, InteractionMap, accepted by CVPR 2025 ! This study improves previous map vectorization methods by fully leveraging local-to-global information interaction in both time and space. (Project , Code, Paper)Jun. 8th, 2024: 💡 The Innovation-Award of the Mapless Driving-Challenge goes to our LGMap Solution ! (Tech-Report , Video)Jun. 1st, 2024: 🏆 Our team win the Championship of the CVPR24 Mapless Driving Challenge ! (Leaderboard)
Vectorized high-definition (HD) maps are essential for
an autonomous driving system. We introduce a novel online mapping
pipeline LGmap, which adept at long-range temporal model.
Firstly, we propose symmetric view transformation(SVT),
a hybrid view transformation module.
Our approach overcomes the limitations of
forward sparse feature representation and utilizing depth perception and SD prior information.
Secondly, we propose hierarchical temporal fusion(HTF)
module. It employs temporal information from local to global,
which empowers the construction of long-range HD map
with high stability. Lastly, we propose a novel ped-crossing
resampling. The simplified ped crossing representation accelerates the instance attention based decoder convergence performance.
Our method achieves 0.66 UniScore in the Mapless Driving OpenLaneV2 test set.
- We propose LGmap, which improves previous map vectorization methods by fully leveraging local-to-global information interaction in both time and space.
- The overall model architecture of LGmap. The entire model is consists of mainly six components: a image backbone equipped with SVT(Symmetric View Transformation), a hierarchical temporal fusion(HTF) module, a unified instance detection and segmentation predictor, a traffic elements detector(YOLO), a Lane-Lane Topology(LLT) and a Lane-TE Topology(LTT)
- Comparison with SOTA methods, LGmap is the 1st-place solution of the CVPR24 *Mapless Driving Challenge, which achieves 0.66 UniScore.
- We visualize results of LGMap in sequential frames. (Demo)
@article{wu2024lgmap,
title={LGmap: Local-to-Global Mapping Network for Online Long-Range Vectorized HD Map Construction},
author={Wu, Kuang and Nian, Sulei and Shen, Can and Yang, Chuan and Li, Zhanbin},
journal={arXiv preprint arXiv:2406.13988},
year={2024}
}