HeightMapNet: Explicit Height Modeling for End-to-End HD Map Learning

TL;DR

HeightMapNet is a novel framework that improves HD map learning from surround-view images by explicitly modeling road surface height, integrating height priors, and utilizing multi-scale BEV features for enhanced accuracy.

Contribution

It introduces HeightMapNet, which explicitly models height information and employs a foreground-background separation network for precise HD map extraction from images.

Findings

Outperforms existing methods on nuScenes and Argoverse 2 datasets.

Effectively models road surface height for improved BEV feature accuracy.

Utilizes multi-scale features to enhance spatial information in HD map learning.

Abstract

Recent advances in high-definition (HD) map construction from surround-view images have highlighted their cost-effectiveness in deployment. However, prevailing techniques often fall short in accurately extracting and utilizing road features, as well as in the implementation of view transformation. In response, we introduce HeightMapNet, a novel framework that establishes a dynamic relationship between image features and road surface height distributions. By integrating height priors, our approach refines the accuracy of Bird's-Eye-View (BEV) features beyond conventional methods. HeightMapNet also introduces a foreground-background separation network that sharply distinguishes between critical road elements and extraneous background components, enabling precise focus on detailed road micro-features. Additionally, our method leverages multi-scale features within the BEV space, optimally utilizing spatial geometric information to boost model performance. HeightMapNet has shown exceptional results on the challenging nuScenes and Argoverse 2 datasets, outperforming several widely recognized approaches. The code will be available at \url{https://github.com/adasfag/HeightMapNet/}.