Painting completion with generative translation models

Image completion has always been an important research area of image processing. With the continuous development of the deep learning model in recent years, further progress has been made in the repair of images. In this paper, we focused on realistic and painting portrait data, studied on semantic inpainting techniques based on regional completions, and proposed an improved generative translation model. Through the context generation network and the image discriminator network, a patch image is generated which should keep consistency between the hole and the surrounding area. Then the completed part will be processed according to the scene structure of the image through the style translation network to ensure the consistency between the generated area and the whole image, which means the repair part can better adapt to the style, texture, and structure of the artistic work. Experiments have shown that our method could achieve good results in the completion of realistic and painting portraits, and it also provided some reference for restoration and identification of art works.