MGMapNet: Multi-Granularity Representation Learning for End-to-End Vectorized HD Map Construction

Jing Yang,Minyue Jiang,Sen Yang,Xiao Tan,Yingying Li,Errui Ding,Hanli Wang,Jingdong Wang

2024-10-10

Abstract:The construction of Vectorized High-Definition (HD) map typically requires capturing both category and geometry information of map elements. Current state-of-the-art methods often adopt solely either point-level or instance-level representation, overlooking the strong intrinsic relationships between points and instances. In this work, we propose a simple yet efficient framework named MGMapNet (Multi-Granularity Map Network) to model map element with a multi-granularity representation, integrating both coarse-grained instance-level and fine-grained point-level queries. Specifically, these two granularities of queries are generated from the multi-scale bird's eye view (BEV) features using a proposed Multi-Granularity Aggregator. In this module, instance-level query aggregates features over the entire scope covered by an instance, and the point-level query aggregates features locally. Furthermore, a Point Instance Interaction module is designed to encourage information exchange between instance-level and point-level queries. Experimental results demonstrate that the proposed MGMapNet achieves state-of-the-art performance, surpassing MapTRv2 by 5.3 mAP on nuScenes and 4.4 mAP on Argoverse2 respectively.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to capture the category and geometric information of map elements simultaneously in high - definition map (HD Map) construction. Currently, the most advanced methods usually only adopt point - level or instance - level representations, ignoring the strong internal relationships between points and instances. This leads to deficiencies in representing lane relationships, etc., or poor performance in capturing the geometric details of irregular or elongated map elements. To solve these problems, the authors propose a new framework - MGMapNet (Multi - Granularity Map Network), which realizes the multi - granularity representation of map elements by integrating coarse - grained instance - level queries and fine - grained point - level queries, thereby capturing the overall and local information of map elements more effectively.

Specifically, MGMapNet solves the problem in the following ways:
1. **Multi - Granularity Aggregator**: Generate instance - level queries and point - level queries from multi - scale bird - eye - view (BEV) features. The instance - level query aggregates features within the entire instance coverage area, while the point - level query aggregates features locally.
2. **Point Instance Interaction**: A module is designed to promote the information exchange between instance - level queries and point - level queries and enhance the internal relationship between them.
3. **Experimental Results**: The experimental results show that MGMapNet has achieved the most advanced performance on both the nuScenes and Argoverse2 datasets, exceeding MapTRv2 by 5.3 mAP and 4.4 mAP respectively.

Through these innovations, MGMapNet is not only more accurate in representing geometric location and category information, but also performs excellently in handling high - definition map construction tasks in complex scenarios.