Research on Automatic Intelligent Coloring of Animation Sketch Based on Enhanced Deep Learning

Abstract

An automatic intelligent coloring model of animation sketch based on enhanced deep learning is proposed. In the proposed model, generative adversarial networks (GANS) are adopted. The U-net network based on the Swish function residual enhancement is used in the generative model, and the ResNet network is used in the discriminant model. The U-net embedded with the Swish Gate module is adopted to transmit feature map information. The perceptual network on the discriminator is used to perceive the perceptual features of the generated image and the actual image and calculate the perceptual loss. Experiment results show that perceptual loss can better capture the difference between black-and-white images and color images, so as to better train the network end-to-end. After comparative analysis, it can be concluded that compared with the existing methods, the proposed model has greater advantages in processing animation sketches. The color images it generates have higher visual quality and richer color diversity and matching.