Joint Eye Tracking and Head Pose Estimation for Gaze Estimation

Ankan Bansal (10095)

Salman Mohammad (10630)

Introduction

Head pose estimation, eye location and gaze estimation from images is an important field of research. They have many applications in commercial, entertainment and strategic sectors. Their uses vary from Human-Computer interaction (e.g. games, pointer control and other vision bases applications) to advertisement research. Driver attention analysis, human behaviour information and control for disabled people are some of the uses of gaze estimation. Some people have used several kinds of restricting hardwares for the purpose of head pose and eye location estimation. So there was a need to develop single static camera based algorithms for this purpose. Head pose and eye location have been separately used for estimation of visual gaze. But these are prone to certain errors and inaccuracies. A combined model of both head pose and eye location estimates can give better results than both.

Report

Project report is available here. Codes can be found in this folder. Final presentation can be downloaded from here. The Boston University head pose dataset is available on this link.