Generating human faces by generative adversarial networks

• Main Author: S Sri Kalki
• Other Authors: Chen Change Loy
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/163273
• Institution: Nanyang Technological University

Video Style transfer is the process of merging the content of one video with the style of another to create a stylized video. In this report, I first study various style transfer techniques such as Adaptive Instance Normalisation (AdaIN), AnimeGAN and GAN N’ Roses. After the various approaches are studied, I then understand the first order motion model of its driving video motion sequences. Finally, study the state-of-the-art StyleGAN and the Toonification algorithm in detail. Furthermore, this report proposes to reimplement state-of-the-art methodologies, investigate the impact of relevant hyperparameters, and offer analysis of these hyperparameters. I expand the existing StyleGAN-based Image Toonification models to Video Toonification. I collect datasets in a total of five styles for the process of style transfer. Finally, I conclude by discussing potential directions for further development.