Strong-Weak Integrated Semi-Supervision for Unsupervised Domain Adaptation

Abstract

Unsupervised domain adaptation (UDA) focuses on transferring knowledge learned in the labeled source domain to the unlabeled target domain. Semi-supervised learning is a proven strategy for improving UDA performance. In this paper, we propose a novel strong-weak integrated semi-supervision (SWISS) learning strategy for unsupervised domain adaptation. Under the proposed SWISS-UDA framework, a strong representative set with high confidence but low diversity target domain samples and a weak representative set with low confidence but high diversity target domain samples are updated constantly during the training process. Both sets are fused randomly to generate an augmented strong-weak training batch with pseudo-labels to train the network during every iteration. Moreover, a novel adversarial logit loss is proposed to reduce the intra-class divergence between source and target domains, which is back-propagated adversarially with a gradient reverse layer between the classifier and the rest of the network. Experimental results based on two popular benchmarks, Office-Home, and DomainNet, show the effectiveness of the proposed SWISS framework with our method achieving the best performance in both Office-Home and DomainNet.