LAUSR

Recently, black-box domain adaptation attracts a lot of attention. It is a new concept to realize domain adaptation of instance image classification with only a cloud API service, reflecting the focus on development of cloud services and concerns about data security. However, the existing black-box domain adaptation methods always only use high-confidence image samples which limits their classification performance. We propose a self-alignment approach to realize black-box domain adaptation. First, an initial model is constructed for target domain. Then, we put target data into source model API to obtain the pseudo-labels and divide the target data into high-confidence and low-confidence parts according to their pseudo-label confidences. By matching the data distributions between these two parts and self-supervised learning on high-confidence part, the classification performance of initial model on both parts samples can be boosted respectively. Information maximization is also applied to further improve classification performance. Experiment results confirm that our method achieves the state-of-the-art performance on image classification.