Skip to main content

Automatic Facial Landmarking

Researcher: Marios Savvides

Research Area: Security of Cyber-Physical Systems

Abstract

Automatic facial landmarking is a key step that enables expression analysis, 3D model generation, pose estimation and even facial recognition. Our center is aiming to improve existing automatic landmarking methods to not only deal with frontal faces, but off-pose ones as well. Our focus is on better modeling facial shape and the texture around a dense set of facial landmarks. Presently our implementation obtains high accuracy when fitting unseen images (with pose variation between -30° and +30°) from outside its training set and outperforms several existing methods.

Outcomes: an advanced automatic facial landmarking system.