Interactive Artistic Multi-style Transfer

Abstract Artistic style transfer is to render an image in the of another image, which a challenge problem both processing and arts. Deep neural networks are adopted artistic achieve remarkable success, such as AdaIN (adaptive instance normalization), WCT (whitening coloring transforms), MST (multimodal transfer), SEMST (structure-emphasized multimodal transfer). These algorithms modify content whole using only one algorithm, easy cause foreground background be blurred together. In this paper, iterative multi-style system built edit with multiple styles by flexible user interaction. First, subjective evaluation experiment art professionals conducted build open framework for transfer, including universal questions personalized answers ten typical styles. Then, we propose interactive system, crop tool designed cut into several parts. For each part, users select algorithm from AdaIN, WCT, MST, referring characteristics summarized experiments. To obtain richer results, provides semantic-based parameter adjustment mode function preserving colors image. Finally, case studies show effectiveness flexibility system.