LabelDistill: Label-guided Cross-modal Knowledge Distillation for Camera-based 3D Object Detection

TL;DR

LabelDistill introduces a novel label-guided cross-modal knowledge distillation method that enhances camera-based 3D object detection by effectively leveraging LiDAR data while mitigating its imperfections, leading to significant performance improvements.

Contribution

The paper proposes a new approach that uses label guidance and feature partitioning to improve cross-modal knowledge transfer in 3D detection, addressing LiDAR measurement ambiguities.

Findings

Improves mAP by 5.1 points over baseline

Enhances NDS by 4.9 points

Demonstrates effective knowledge transfer from LiDAR to camera-based detectors

Abstract

Recent advancements in camera-based 3D object detection have introduced cross-modal knowledge distillation to bridge the performance gap with LiDAR 3D detectors, leveraging the precise geometric information in LiDAR point clouds. However, existing cross-modal knowledge distillation methods tend to overlook the inherent imperfections of LiDAR, such as the ambiguity of measurements on distant or occluded objects, which should not be transferred to the image detector. To mitigate these imperfections in LiDAR teacher, we propose a novel method that leverages aleatoric uncertainty-free features from ground truth labels. In contrast to conventional label guidance approaches, we approximate the inverse function of the teacher's head to effectively embed label inputs into feature space. This approach provides additional accurate guidance alongside LiDAR teacher, thereby boosting the performance of the image detector. Additionally, we introduce feature partitioning, which effectively transfers knowledge from the teacher modality while preserving the distinctive features of the student, thereby maximizing the potential of both modalities. Experimental results demonstrate that our approach improves mAP and NDS by 5.1 points and 4.9 points compared to the baseline model, proving the effectiveness of our approach. The code is available at https://github.com/sanmin0312/LabelDistill