Digital avatar customization using Cycle Generative Adversarial Networks

Abstract

This document presents the Final Degree Work report of the Video Games Design and Development Degree by Lucía González García. The primary aim of this project is to enhance the creation and personalization of digital avatars using Deep Learning techniques, particularly Generative Adversarial Networks (GAN) technology. The objective is to enable users to generate avatars based on their own photographs, such as selfies, and apply various desired image styles. To achieve this, an application will be developed, allowing users to upload their images and apply predefined styles such as manga, cartoons, and more. The outcome will be a harmonious combination of the original image and the selected style, resulting in a coherent visual output.