Efficient Avatar Generation from Arbitrary Images

AR/VR may be the next frontier for online human communications and interactions. The ability to produce photorealistic avatars dramatically improves the feeling of immersion and connection in applications utilizing AR/VR. However, current methods of face capture are time-consuming and involve expensive cameras and sensors. In this project, we explore deep learning methods for generating face avatars using arbitrary images of a subject acquired on inexpensive consumer cameras, such as smartphone selfies, from various viewing angles and at different instances. Furthermore, the attributes of the generated avatars can be edited and animated. The project’s success will provide the partner organization with capabilities allowing it to build a new innovative product in the AR/VR space.

Faculty Supervisor:

Karan Singh

Student:

Partner:

AMD Canada

Discipline:

Computer science

Sector:

Manufacturing; Professional, scientific and technical services

University:

University of Toronto

Program: