Domain Adaptation in 3D Object Detection with Gradual Batch Alternation Training

TL;DR

This paper introduces Gradual Batch Alternation, a training strategy for LiDAR-based 3D object detection that gradually shifts from source to target domain data, improving adaptation performance across multiple autonomous driving datasets.

Contribution

The paper proposes a simple, effective training method for domain adaptation in 3D object detection that gradually reduces source data influence during training.

Findings

Significant performance improvements over prior methods.

Effective across four benchmark autonomous driving datasets.

Demonstrates robustness and adaptability of the proposed strategy.

Abstract

We consider the problem of domain adaptation in LiDAR-based 3D object detection. Towards this, we propose a simple yet effective training strategy called Gradual Batch Alternation that can adapt from a large labeled source domain to an insufficiently labeled target domain. The idea is to initiate the training with the batch of samples from the source and target domain data in an alternate fashion, but then gradually reduce the amount of the source domain data over time as the training progresses. This way the model slowly shifts towards the target domain and eventually better adapt to it. The domain adaptation experiments for 3D object detection on four benchmark autonomous driving datasets, namely ONCE, PandaSet, Waymo, and nuScenes, demonstrate significant performance gains over prior arts and strong baselines.