Enhancing Vectorized Map Perception with Historical Rasterized Maps

TL;DR

This paper proposes HRMapNet, leveraging a low-cost Historical Rasterized Map to enhance online vectorized map perception and integrates it in two state-of-the-art methods, significantly improving their performance on both the nuScenes and Argoverse 2 datasets.

Abstract

In autonomous driving, there is growing interest in end-to-end online vectorized map perception in bird's-eye-view (BEV) space, with an expectation that it could replace traditional high-cost offline high-definition (HD) maps. However, the accuracy and robustness of these methods can be easily compromised in challenging conditions, such as occlusion or adverse weather, when relying only on onboard sensors. In this paper, we propose HRMapNet, leveraging a low-cost Historical Rasterized Map to enhance online vectorized map perception. The historical rasterized map can be easily constructed from past predicted vectorized results and provides valuable complementary information. To fully exploit a historical map, we propose two novel modules to enhance BEV features and map element queries. For BEV features, we employ a feature aggregation module to encode features from both onboard images and the historical map. For map element queries, we design a query initialization module to endow queries with priors from the historical map. The two modules contribute to leveraging map information in online perception. Our HRMapNet can be integrated with most online vectorized map perception methods. We integrate it in two state-of-the-art methods, significantly improving their performance on both the nuScenes and Argoverse 2 datasets. The source code is released at https://github.com/HXMap/HRMapNet.

Figures (8)

• Figure 1: Pipeline of the proposed HRMapNet. The words in red are what we design to maintain and leverage a historical map for online perception.
• Figure 2: Architecture of the proposed HRMapNet. The gray blocks are kept unchanged from SOTA online vectorized map perception methods.
• Figure 3: Visualized results in three challenging scenarios: severe occlusion, rainy day and poor lighting at night. All three methods are trained with 24 epochs. Ours is based on MapTRv2. Lane divider, pedestrian crossing and road boundary are illustrated in red, green and blue respectively.
• Figure 4: Visualized results. The last column is the retrieved local rasterized maps in HRMapNet. Lane divider, pedestrian crossing and road boundary are illustrated in red, green and blue respectively.
• Figure 5: Visualized results. The last column is the retrieved local rasterized maps in HRMapNet. Lane divider, pedestrian crossing and road boundary are illustrated in red, green and blue respectively.