Uncertainty-aware Mean Teacher for Source-free Unsupervised Domain Adaptive 3D Object Detection
This addresses domain adaptation challenges in 3D object detection for autonomous driving, but it is incremental as it builds on existing pseudo-label and mean teacher methods.
The paper tackles the problem of incorrect pseudo-labels in source-free unsupervised domain adaptation for 3D object detection by proposing an uncertainty-aware mean teacher framework that implicitly filters errors during training. It achieves state-of-the-art performance on cross-dataset and cross-weather scenarios on the KITTI lidar dataset.
Pseudo-label based self training approaches are a popular method for source-free unsupervised domain adaptation. However, their efficacy depends on the quality of the labels generated by the source trained model. These labels may be incorrect with high confidence, rendering thresholding methods ineffective. In order to avoid reinforcing errors caused by label noise, we propose an uncertainty-aware mean teacher framework which implicitly filters incorrect pseudo-labels during training. Leveraging model uncertainty allows the mean teacher network to perform implicit filtering by down-weighing losses corresponding uncertain pseudo-labels. Effectively, we perform automatic soft-sampling of pseudo-labeled data while aligning predictions from the student and teacher networks. We demonstrate our method on several domain adaptation scenarios, from cross-dataset to cross-weather conditions, and achieve state-of-the-art performance in these cases, on the KITTI lidar target dataset.