# MapSeg: Segmentation guided structured model for online HD map construction

arXiv: https://arxiv.org/abs/2311.02503 · HTML: https://mlsanity.com/paper/2311.02503/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 6 of its 6 references that are in the map.
- First posted on arXiv 13 Jun 2023.

## Builds on (most cited first, 6 of 6)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation](https://mlsanity.com/paper/1802.02611/) (2018)
- [Decoupled Weight Decay Regularization](https://mlsanity.com/paper/1711.05101/) (2017)
- [MapTR: Structured Modeling and Learning for Online Vectorized HD Map Construction](https://mlsanity.com/paper/2208.14437/) (2022)
- [VectorMapNet: End-to-end Vectorized HD Map Learning](https://mlsanity.com/paper/2206.08920/) (2022)

## Abstract (arXiv)

The development of online high-definition maps is significant since they provide real-time, accurate, and updatable geographic information for location-based applications, such as autonomous driving and intelligent transportation, thus improving the performance and reliability of these applications. Previous works, such as VectorMapNet and MapTR, show that direct model generation of vectorized HD maps is a promising solution. However, these methods did not take into account the usage of global semantic information to improve map construction accuracy. To address this limitation, we propose a segmentation-guided structured model (MapSeg) for online HD map construction. Specifically, we added a UV segmentation module (USM) and a BEV segmentation module (BSM) based on the MapTR structure, enabling the model to better capture the semantic information.