Anime sketch coloring with swish-gated residual U-net

Abstract

Anime sketch coloring is to fill the color into the anime sketches to obtain the colorful anime images and it is a new research direction in deep learning technology. Currently, generative adversarial networks (GANs) have been used for anime sketch coloring and achieved some results. However, the colorful images generated by the anime sketch coloring methods based on GANs generally have poor coloring effects. In this paper, an efficient anime sketch coloring method based on swish-gated residual U-net (SGRU) is proposed to solve the above problems. In SGRU, the proposed swish layer and swish-gated residual blocks (SGRBs) effectively filter the information transmitted by each level and speed up the convergence of the network. The perceptual loss and the per-pixel loss are used to constitute the final loss of SGRU. The final loss function reflects the coloring results more realistically and can control the effect of coloring more effectively. SGRU can automatically color the sketch without providing any coloring hints in advance and can be trained end-to-end with the sketch and the corresponding color image. Experimental results show that our method performs better than other state-of-the-art coloring methods, and can achieve the colorful images with higher visual quality.