SemanticBEVFusion: Rethink LiDAR-Camera Fusion in Unified Bird's-Eye View Representation for 3D Object Detection
This work addresses a key challenge in autonomous driving by enhancing 3D object detection through improved sensor fusion, though it appears incremental as it builds on existing bird's-eye view fusion strategies.
The paper tackled the problem of effectively fusing LiDAR and camera data for 3D object detection in autonomous driving by introducing SemanticBEVFusion, which deeply integrates features in a unified bird's-eye view representation, achieving state-of-the-art performance on the nuScenes dataset with notable improvements for distant objects.
LiDAR and camera are two essential sensors for 3D object detection in autonomous driving. LiDAR provides accurate and reliable 3D geometry information while the camera provides rich texture with color. Despite the increasing popularity of fusing these two complementary sensors, the challenge remains in how to effectively fuse 3D LiDAR point cloud with 2D camera images. Recent methods focus on point-level fusion which paints the LiDAR point cloud with camera features in the perspective view or bird's-eye view (BEV)-level fusion which unifies multi-modality features in the BEV representation. In this paper, we rethink these previous fusion strategies and analyze their information loss and influences on geometric and semantic features. We present SemanticBEVFusion to deeply fuse camera features with LiDAR features in a unified BEV representation while maintaining per-modality strengths for 3D object detection. Our method achieves state-of-the-art performance on the large-scale nuScenes dataset, especially for challenging distant objects. The code will be made publicly available.