Conditional Adaptation Deep Networks for Unsupervised Cross Domain Image Classifcation

Unsupervised domain adaptation aims at improving the performance on an unknown target domain by utilizing the knowledge learned from a related source domain. In this paper, we proposed a conditional adaptation deep network (CADN) to deal with the unsupervised domain adaptation problem. CADN is implemented based on a deep neural network and attempts to learn domain invariant features. We present a conditional adaptation strategy to reduce the domain distribution discrepancy and to solve the category mismatch and class prior bias, which are usually ignored in marginal adaptation approaches. We perform a set of comparative experiments on the standard domain adaptation benchmarks. The results demonstrate that the proposed CADN outperforms previous adaptation methods.