Gaze Data Collection with the Off-the-Shelf Devices

摘要

Gaze is a very important modality in Human-computer Interaction (HCI). Sufficient gaze data is one of the vital foundations for automatic gaze tracking. In this paper, we propose a method to collect automatic labelled gaze data with the off-the-shelf devices. We use motion trackers to locate head pose and gaze orientation. The observed cursor on a calibrated screen moves with the pre-defined patterns. The captured sequences are then automatic labelled. The expanding database has 23676 samples of twelve persons. Our method is simple and reliable, and our system can be set up at any condition with the off-the-shelf devices. The ground truth data is useful for further research on gaze and head pose.