Self-supervised Learning of 3D Facial Reconstruction from Videos

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Defense

Title: "Self-supervised Learning of 3D Facial Reconstruction from Videos"

by

LIU Zichen

Abstract:

3D facial reconstruction is the task of getting the 3D facial model of a 
human. It has been an extensively studied topic in the computer vision and 
graphics community and has many applications in VR/AR telepresence and 
games. However, annotated 3D labeled facial data is expensive to get. As a 
result, some unsupervised methods that do not rely on labeled data have 
been proposed for 3D facial reconstruction. Recently, methods that can 
create animatable personal facial avatars have drawn much attention and 
shown promising results in readily available settings (e.g., from 
monocular videos or portrait photo sequences). The goal of this final year 
thesis is creating personalized facial avatars under fine-grained control 
over expression while approaching photorealism and acquiring accurate 
facial geometry.


Date            : 2 May 2023 (Tuesday)

Time            : 10:00 - 10:40

Venue           : Room 2406 (near lifts 17/18), HKUST

Advisor         : Dr. XU Dan

2nd Reader      : Dr. WANG Shuai