BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation

TL;DR

BEVFusion introduces a unified bird's-eye view framework for multi-sensor fusion in autonomous driving, significantly improving semantic and geometric perception accuracy while reducing computational latency.

Contribution

It proposes a task-agnostic, multi-task fusion method that unifies multi-modal features in BEV space, overcoming limitations of point-level fusion and optimizing view transformation efficiency.

Findings

Achieves state-of-the-art results on nuScenes with higher accuracy.

Reduces view transformation latency by over 40x.

Supports multiple perception tasks with minimal architectural changes.

Abstract

Multi-sensor fusion is essential for an accurate and reliable autonomous driving system. Recent approaches are based on point-level fusion: augmenting the LiDAR point cloud with camera features. However, the camera-to-LiDAR projection throws away the semantic density of camera features, hindering the effectiveness of such methods, especially for semantic-oriented tasks (such as 3D scene segmentation). In this paper, we break this deeply-rooted convention with BEVFusion, an efficient and generic multi-task multi-sensor fusion framework. It unifies multi-modal features in the shared bird's-eye view (BEV) representation space, which nicely preserves both geometric and semantic information. To achieve this, we diagnose and lift key efficiency bottlenecks in the view transformation with optimized BEV pooling, reducing latency by more than 40x. BEVFusion is fundamentally task-agnostic and seamlessly supports different 3D perception tasks with almost no architectural changes. It establishes the new state of the art on nuScenes, achieving 1.3% higher mAP and NDS on 3D object detection and 13.6% higher mIoU on BEV map segmentation, with 1.9x lower computation cost. Code to reproduce our results is available at https://github.com/mit-han-lab/bevfusion.