Style Transfer 기반 Unsupervised Domain Adaptation

Abstract

When we perform a recognition task using deep neural networks, the recognition rate drops because of the difference between the distribution of training data and test data. To solve this problem, we propose a domain adaptation method using a style transfer method. In our method, the style transfer is performed to the target data so that it looks like the source domain. The transferred images are used for the input to the deep neural network trained by the image of the source domain. Finally, we show that our proposed method is effective to improve the recognition rate.