ArXiv Preprint (arXiv 2304.09807)
Aug. 30th, 2023: We release an initial version of VMA.Aug. 9th, 2023: Code will be released in around 3 weeks.
demo.mp4
TL;DR VMA is a general map auto annotation framework based on MapTR, with high flexibility in terms of spatial scale and element type.
- Installation
- Prepare Dataset
- Inference on SD data (we only provide some samples of SD data for inference, since SD data is owned by Horizon)
- Train and Eval on NYC data
Remote sensing:
Urban scene:
Highway scene:
If you find VMA is useful in your research or applications, please consider giving us a star π and citing it by the following BibTeX entry.
@inproceedings{VMA,
title={VMA: Divide-and-Conquer Vectorized Map Annotation System for Large-Scale Driving Scene},
author={Chen, Shaoyu and Zhang, Yunchi and Liao, Bencheng, Xie, Jiafeng and Cheng, Tianheng and Sui, Wei and Zhang, Qian and Liu, Wenyu and Huang, Chang and Wang, Xinggang},
booktitle={arXiv preprint arXiv:2304.09807},
year={2023}
}