Deep Learning-Based Creative Intention Understanding and Color Suggestions for Illustration

Abstract

With the gradually maturity of deep learning and model training, machine learning is increasingly used in image processing, including style transfer, image repair, image generation, etc. In these studies, although artificial intelligence could accurately reproduce different artistic styles and be able to generate realistic images, illustrators’ creative experiences were ignored. Machine replaces almost all the work of human in these applications. But human’s desire for the painting experience and ability will not disappear. We believe that based on learning the creative intentions of illustrators, machine can make suggestions for illustrators’ problems and help them improve their capabilities. It will be a more harmonious cooperation direction for human and artificial intelligence in illustration field. This paper takes color suggestion as an example. We analyze the difficulties and needs of illustrators when they color the paintings. And we propose a method of using machine learning to assist illustrators in improving their coloring ability. Based on the color of the input works from illustrators, it will optimize the choice of colors, the arrangement and proportion of different colors in canvas to help illustrators understand their weakness in coloring and improvement directions visually.