Explore the LiDAR-Camera Dynamic Adjustment Fusion for 3D Object Detection

TL;DR

This paper introduces a dynamic adjustment fusion method for LiDAR and camera data to improve 3D object detection in autonomous driving, addressing modality gaps through domain alignment and adaptive learning.

Contribution

It proposes a triphase domain aligning module and adaptive learning techniques to enhance multimodal fusion for 3D detection, which is a novel approach in this context.

Findings

Achieves competitive results on nuScenes dataset

Effectively reduces modality distribution gaps

Improves 3D detection accuracy

Abstract

Camera and LiDAR serve as informative sensors for accurate and robust autonomous driving systems. However, these sensors often exhibit heterogeneous natures, resulting in distributional modality gaps that present significant challenges for fusion. To address this, a robust fusion technique is crucial, particularly for enhancing 3D object detection. In this paper, we introduce a dynamic adjustment technology aimed at aligning modal distributions and learning effective modality representations to enhance the fusion process. Specifically, we propose a triphase domain aligning module. This module adjusts the feature distributions from both the camera and LiDAR, bringing them closer to the ground truth domain and minimizing differences. Additionally, we explore improved representation acquisition methods for dynamic fusion, which includes modal interaction and specialty enhancement. Finally, an adaptive learning technique that merges the semantics and geometry information for dynamical instance optimization. Extensive experiments in the nuScenes dataset present competitive performance with state-of-the-art approaches. Our code will be released in the future.