Recovering 3-D gaze scan path and scene structure from inside-out camera

Abstract

First-Person Vision (FPV) is a wearable sensor that takes images from a user's visual field and interprets them, with available information about the user's head motion and gaze, through eye tracking [1]. Measuring the 3-D gaze trajectory of a user moving dynamically in 3-D space is interesting for understanding a user's intention and behavior. In this paper, we present a system for recovering 3-D scan path and scene structure in 3-D space on the basis of ego-motion computed from an inside-out camera. Experimental results show that the 3-D scan paths of a user moving in complex dynamic environments were recovered.