SRoUDA: Meta Self-Training for Robust Unsupervised Domain Adaptation

As acquiring manual labels on data could be costly, unsupervised domain adaptation (UDA), which transfers knowledge learned from a rich-label dataset to the unlabeled target dataset, is gaining increasingly more popularity. While extensive studies have been devoted improving model accuracy domain, an important issue of robustness neglected. To make things worse, conventional adversarial training (AT) methods for are inapplicable under UDA scenario since they train models examples that generated by supervised loss function. In this paper, we present new meta self-training pipeline, named SRoUDA, models. Based paradigm, SRoUDA starts with pre-training source applying baseline labeled and taraget developed random masked augmentation (RMA), then alternates between pseudo-labeled fine-tuning step. allows direct incorporation AT in UDA, step further helps mitigating error propagation noisy pseudo labels. Extensive experiments various benchmark datasets demonstrate state-of-the-art performance where it achieves significant improvement without harming clean accuracy.