Facial Landmark Detection with Synthetic data

Facial landmark detection is a computer vision problem where the goal is to predict the location of specific points on a face, like the eyes, nose, and mouth. This is useful for things like facial recognition and 3D modeling. To train a model to do this, we need a lot of images with those points already marked, which can be expensive and time-consuming. So instead, we can create synthetic images using scans of real people’s faces. Models trained on synthetic data have shown promising outcomes and have been successful. However, these models don’t work well on images taken with helmet-mounted cameras, which are commonly used in film and video games. This research aims to create synthetic data that looks like it was taken with these cameras and train models on it to improve performance.

Faculty Supervisor:

Steve Engels

Student:

Partner:

Ubisoft Toronto

Discipline:

Computer science

Sector:

Information and cultural industries

University:

University of Toronto

Program: