Detecting As Labeling: Rethinking LiDAR-camera Fusion in 3D Object Detection

TL;DR

This paper introduces DAL, a novel LiDAR-camera fusion paradigm for 3D object detection that reduces overfitting and improves performance by mimicking data annotation, offering a simple, portable, and effective baseline.

Contribution

The paper proposes DAL, a new paradigm that imitates data annotation for LiDAR-camera fusion, significantly enhancing detection performance and robustness.

Findings

DAL achieves superior speed-accuracy trade-off.

DAL substantially improves detection performance.

The method is simple, portable, and effective.

Abstract

3D object Detection with LiDAR-camera encounters overfitting in algorithm development which is derived from the violation of some fundamental rules. We refer to the data annotation in dataset construction for theory complementing and argue that the regression task prediction should not involve the feature from the camera branch. By following the cutting-edge perspective of 'Detecting As Labeling', we propose a novel paradigm dubbed DAL. With the most classical elementary algorithms, a simple predicting pipeline is constructed by imitating the data annotation process. Then we train it in the simplest way to minimize its dependency and strengthen its portability. Though simple in construction and training, the proposed DAL paradigm not only substantially pushes the performance boundary but also provides a superior trade-off between speed and accuracy among all existing methods. With comprehensive superiority, DAL is an ideal baseline for both future work development and practical deployment. The code has been released to facilitate future work on https://github.com/HuangJunJie2017/BEVDet.