DistillBEV: Boosting Multi-Camera 3D Object Detection with Cross-Modal Knowledge Distillation

TL;DR

This paper introduces DistillBEV, a method that enhances multi-camera BEV 3D object detection by distilling knowledge from LiDAR-based models, significantly improving performance on the nuScenes benchmark.

Contribution

The paper presents a novel cross-modal knowledge distillation approach that effectively transfers geometric information from LiDAR to camera-based models for improved 3D detection.

Findings

Significant performance gains on nuScenes benchmark.

Effective feature imitation from LiDAR to camera models.

Enhanced multi-scale and temporal feature transfer.

Abstract

3D perception based on the representations learned from multi-camera bird's-eye-view (BEV) is trending as cameras are cost-effective for mass production in autonomous driving industry. However, there exists a distinct performance gap between multi-camera BEV and LiDAR based 3D object detection. One key reason is that LiDAR captures accurate depth and other geometry measurements, while it is notoriously challenging to infer such 3D information from merely image input. In this work, we propose to boost the representation learning of a multi-camera BEV based student detector by training it to imitate the features of a well-trained LiDAR based teacher detector. We propose effective balancing strategy to enforce the student to focus on learning the crucial features from the teacher, and generalize knowledge transfer to multi-scale layers with temporal fusion. We conduct extensive evaluations on multiple representative models of multi-camera BEV. Experiments reveal that our approach renders significant improvement over the student models, leading to the state-of-the-art performance on the popular benchmark nuScenes.