Immersive Traditional Chinese Portrait Painting: Research on Style Transfer and Face Replacement

Traditional Chinese portrait is popular all over the world because of its unique oriental charm. However, how to use neural network to express the aesthetic and feelings in instantiated Chinese portrait effectively is still a challenging problem. This paper proposes a Photo to Chinese Portrait method (P-CP) providing immersive traditional Chinese portrait painting experience. Our method can produce two groups intriguing pictures. One is Chinese Portrait Style Picture, the other is Immersive Chinese Portrait Picture. We pay attention to neural style transfer for traditional Chinese portrait for the first time, and have trained a fast feedforward generative network to extract the corresponding style. The generative network principle is explored to guide the style transfer adjustment in detail. Face replacement is added to form a more appealing stylized effect. We also solve the problems of color and light violation, and unnatural seam. We hope this work offers a deeper and immersive conversation between modern society and antiques, and provides a useful step towards related interdisciplinary areas.