Generating Anime from Real Human Image with Adversarial Training

With growing interest in animation movie, the demand for building an automated system to convert the real video into animation is higher than ever. The frame-by-frame editing of animation generation process is very costly and time-consuming. To help to generate animation quickly and easily in an automated process we proposed a generative model that converts real-world images into corresponding-animation image without losing important details of the source image. We used a variation of the generative adversarial network as underlying architecture with custom loss function to preserve the content of source image which converting to animation image.