BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers

TL;DR

BEVFormer introduces a novel spatiotemporal transformer framework that learns unified bird's-eye-view representations from multi-camera images, significantly improving 3D perception accuracy for autonomous driving tasks.

Contribution

It is the first to integrate spatial and temporal attention mechanisms in a unified BEV framework for multi-camera perception.

Findings

Achieves 56.9% NDS on nuScenes test set, surpassing previous methods.

Improves velocity estimation accuracy and object recall in low visibility conditions.

Performs on par with LiDAR-based methods without using LiDAR data.

Abstract

3D visual perception tasks, including 3D detection and map segmentation based on multi-camera images, are essential for autonomous driving systems. In this work, we present a new framework termed BEVFormer, which learns unified BEV representations with spatiotemporal transformers to support multiple autonomous driving perception tasks. In a nutshell, BEVFormer exploits both spatial and temporal information by interacting with spatial and temporal space through predefined grid-shaped BEV queries. To aggregate spatial information, we design spatial cross-attention that each BEV query extracts the spatial features from the regions of interest across camera views. For temporal information, we propose temporal self-attention to recurrently fuse the history BEV information. Our approach achieves the new state-of-the-art 56.9\% in terms of NDS metric on the nuScenes \texttt{test} set, which is 9.0 points higher than previous best arts and on par with the performance of LiDAR-based baselines. We further show that BEVFormer remarkably improves the accuracy of velocity estimation and recall of objects under low visibility conditions. The code is available at \url{https://github.com/zhiqi-li/BEVFormer}.