An Empirical Study of Pseudo-Labeling for Image-based 3D Object Detection

TL;DR

This paper investigates the use of pseudo-labeling as a semi-supervised approach to improve image-based 3D object detection, demonstrating significant performance gains and surprising findings that pseudo-labels can outperform ground-truth annotations in certain settings.

Contribution

It provides extensive experimental analysis of pseudo-labeling effectiveness for image-based 3D detection, revealing its potential as a cost-effective alternative to manual annotations.

Findings

Pseudo-labeling improves detection performance significantly.

Models trained with pseudo-labels can outperform those trained with ground-truth.

Achieved 20.23 AP on KITTI-3D, surpassing baseline by 6.03 AP.

Abstract

Image-based 3D detection is an indispensable component of the perception system for autonomous driving. However, it still suffers from the unsatisfying performance, one of the main reasons for which is the limited training data. Unfortunately, annotating the objects in the 3D space is extremely time/resource-consuming, which makes it hard to extend the training set arbitrarily. In this work, we focus on the semi-supervised manner and explore the feasibility of a cheaper alternative, i.e. pseudo-labeling, to leverage the unlabeled data. For this purpose, we conduct extensive experiments to investigate whether the pseudo-labels can provide effective supervision for the baseline models under varying settings. The experimental results not only demonstrate the effectiveness of the pseudo-labeling mechanism for image-based 3D detection (e.g. under monocular setting, we achieve 20.23 AP for moderate level on the KITTI-3D testing set without bells and whistles, improving the baseline model by 6.03 AP), but also show several interesting and noteworthy findings (e.g. the models trained with pseudo-labels perform better than that trained with ground-truth annotations based on the same training data). We hope this work can provide insights for the image-based 3D detection community under a semi-supervised setting. The codes, pseudo-labels, and pre-trained models will be publicly available.