Digital makeup using deep learning methods

Abstract

Makeup transfer algorithm is extensively used worldwide in technology. The purpose of makeup transfer is to extract and transform the makeup style from various makeup images to raw non-makeup image. It is similar to physical make up as it begins with makeup base and ends in skin and colour make up while preserving the face identities, so that the users are able to try makeup virtually and find more suitable makeup style on their faces. With development in makeup transfer, new approaches are introduced such as generative adversarial network. This project includes research on facial parsing, BeautyGAN and DMT for digital make up and conducts experiments using pre-trained models and CelebAMask-HQ dataset to compare the results and find better solutions.