Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Wenzhao Qiu, Shanmin Pang, Hao zhang, Jianwu Fang, Jianru Xue

TL;DR

The HeightMap-Net framework is introduced, a novel framework that establishes a dynamic relationship between image features and road surface height distributions and refines the accuracy of Bird's-Eye-View (BEV) features beyond conventional methods.

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/}.

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

TL;DR

The HeightMap-Net framework is introduced, a novel framework that establishes a dynamic relationship between image features and road surface height distributions and refines the accuracy of Bird's-Eye-View (BEV) features beyond conventional methods.

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/}.

Paper Structure

This paper contains 20 sections, 8 equations, 4 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Overview
  5. Foreground-Background Separation
  6. Height Prediction Mechanism
  7. Multi-Scale Feature Fusion
  8. Training Loss
  9. Experiment
  10. Experimental Setup
  11. Implementation Details
  12. Main Results
  13. Results on the nuScenes $\texttt{val}$ Set
  14. Results on the nuScenes $\texttt{Test}$ Set
  15. Results on the Argoverse 2 $\texttt{Val}$ Set
  16. ...and 5 more sections

Figures (4)

  • Figure 1: The overview of our proposed HeightMapNet. Our method consists of three main components: a foreground-background separation network that emphasizes roadway features, a height prediction mechanism for PV-to-BEV transformation, and a multi-scale feature fusion module that enhances BEV representation.
  • Figure 2: Illustration of height prediction mechanism.
  • Figure 3: Illustration of multi-scale feature fusion.
  • Figure 4: Visualizations on the nuScenes $\texttt{val}$ set under different weather conditions.