Domain Adaptation in 3D Object Detection with Gradual Batch Alternation Training
This addresses the problem of adapting 3D object detection models from labeled source domains to insufficiently labeled target domains in autonomous driving, representing an incremental improvement over prior methods.
The paper tackles domain adaptation in LiDAR-based 3D object detection by proposing Gradual Batch Alternation Training, which gradually reduces source domain data during training to shift the model toward the target domain, achieving significant performance gains on four autonomous driving datasets.
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.