Fast Style Transfer for Chinese Tranditional Ink Painting

摘要

By style transfer mechanism, the style of another image can be illustrated toward another target picture. However, calculating speed and quality are two tricky issues. In this paper, we propose an accelerated version of the transfer for Chinese traditional ink painting style. By reducing the size of the rendering network, the training speed can be increased to 1.27 times. It is about the similar quality as the image generated by the original model. This can also reflect the characteristics of ink painting: light ink, black and white.