DataBase Generation for Eye Detection with Spectacles

Abstract

Face and subsequently eye detection is primary step for facial image processing. Most of these applications faces are considered without spectacles that poses challenge to detect ocular features. In this work, we aim to create a facial data-base with and without spectacles to facilitate the validation of algorithms for face or eye detection. We designed an experiment to incorporate variability in several parameters to be detected and measured. We have generated the ground truth to provide a platform for the researchers to quantitatively evaluate the performance of their algorithms on this database. Eye movement in terms of view angle and velocity are measured and co-related with the stimulus movement on the screen empirically.