Towards an Artificial Intelligence Aided Design Approach: Application to Anime Faces with Generative Adversarial Networks

Abstract Ever since the inception of Machine Learning and Artificial Intelligence, the basic motto for most of research works has been to bring the machines at par with human intelligence. Designing new products and artifacts is one of the many fields where it is very difficult to enable computing machines replicate human creativity and innovativeness. Design processes in engineering fields as well as in arts follow methodical series of steps to create new products. Due to high demands of customized products and services, competitors tend to shorten the time-to-market periods, using advanced Computer-Aided Design programs. These programs play important roles to assist designers in digitizing blueprints and automating repetitive tasks. However, they fail to boost designer creativity by generating or suggesting new ideas or designs based on existing products or their variants. In order to boost creativity in the entertainment industry, we propose in this paper a new approach based on unsupervised learning techniques to create variants of a given artifact or product blueprints. Within the field of designing new cartoon characters, our proposed approach relies on Generative Adversarial Neural Networks [1] to create new anime or cartoon faces on their own without any human intervention. It learns features and characteristics from an image training dataset and combines them to create new features and thus builds a new image which is not present in the training dataset. This applied approach attempts to not only help artists and designers to have a preview of the possible new and unique avatars but also would prevent any copyright infringements.