Creative Approaches Towards Development of Computing and Multidisciplinary IT Solutions for Society

12Creating a Video from Facial Image Using Conditional Generative Adversarial Network

Bui Thanh Hung^*, Ho Vo Hoang Duy and Vo Quoc Huy

Data Science Laboratory, Faculty of Information Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Vietnam

Abstract

Create a video from facial images that holds significance in generating natural-looking videos from a single image. This technique is widely used in various fields such as filmmaking and social media. Previous methods had had limitations, such as creating high-quality videos lacking naturalness in reproducing character movements. Some studies have focused solely on producing high-quality videos, resulting in a loss of diversity in the content and style of images. In this study, we propose a method for creating a short video with natural facial movements of the lips, eyes, and related facial parts using deep learning techniques, convolutional neural networks (CNN), Hidden Affine transformation combined with Conditional Generative Adversarial Network (cGAN), image processing techniques, and computer vision methods. We evaluated this method on CK-Mixed datasets and compared it with other methods. Based on these results, we will develop an application that can create a facial motion video from a single input image and test its practicality.

Keywords: Creating a video, facial images, deep learning, CNN, hidden affine transformation, cGAN

12.1 Introduction

In recent years, significant advances have been made in ...