Pseudo-labeling for Scalable 3D Object Detection

TL;DR

This paper demonstrates that pseudo-labeling can significantly improve 3D object detection in autonomous vehicles by leveraging unlabeled data, leading to better performance and domain adaptation with simpler models.

Contribution

It introduces pseudo-labeling for 3D detection, showing it enhances accuracy, reduces reliance on labeled data, and improves domain generalization across architectures.

Findings

Pseudo-label-trained models outperform supervised models trained on 3-10 times more data.

Using PointPillars, state-of-the-art accuracy is achieved with pseudo-labeling.

Pseudo-labeling improves domain generalization to unlabeled target environments.

Abstract

To safely deploy autonomous vehicles, onboard perception systems must work reliably at high accuracy across a diverse set of environments and geographies. One of the most common techniques to improve the efficacy of such systems in new domains involves collecting large labeled datasets, but such datasets can be extremely costly to obtain, especially if each new deployment geography requires additional data with expensive 3D bounding box annotations. We demonstrate that pseudo-labeling for 3D object detection is an effective way to exploit less expensive and more widely available unlabeled data, and can lead to performance gains across various architectures, data augmentation strategies, and sizes of the labeled dataset. Overall, we show that better teacher models lead to better student models, and that we can distill expensive teachers into efficient, simple students. Specifically, we demonstrate that pseudo-label-trained student models can outperform supervised models trained on 3-10 times the amount of labeled examples. Using PointPillars [24], a two-year-old architecture, as our student model, we are able to achieve state of the art accuracy simply by leveraging large quantities of pseudo-labeled data. Lastly, we show that these student models generalize better than supervised models to a new domain in which we only have unlabeled data, making pseudo-label training an effective form of unsupervised domain adaptation.