A source free domain adaptation model based on adversarial learning for image classification

The unsupervised domain adaptive classification task can learn domain-invariant features between the unlabeled target domain and the labeled source domain, thereby improving the classification performance in target domain. However, privacy protection and memory constrains often make it difficult to obtain labeled source domain samples, which become bottlenecks for the traditional domain adaptation. To this end, we propose a novel source free domain adaptive classification model. This model helps us to obtain a classifier with better effect in the target domain only by using the classifier trained in source domain and the target domain data without any source domain data. Firstly we propose a novel conditional information generative adversarial module based on combined discriminators. By confronting between combined discriminators and the generator, the middle domain with pseudo-labels is generated to solve the problem of missing source domain. Then when training the new classifier in domain adaptation module, we add a distillation loss mechanism to deal with the lack of source domain data supervision, thereby minimizing the difference between the old classifier response and the new classifier response to ensure that the network output retains the source domain information. Three groups of 10 datasets are used to verify this models. The results show that our methods can effectively solve the problem of source free domain adaptive classification and improve the classification accuracy in each domain.