Recent Progress on Eye-Tracking and Gaze Estimation for AR/VR Applications: A Review

Abstract

Visual information is crucial in human life, not only providing critical support for communication, learning, and decision-making, but also playing a key role in psychology, medicine, and science. Eye-tracking and gaze estimation have promoted the development of foveated rendering in wearable virtual reality and augmented reality glasses. This review summarizes the recent development on gaze estimation and discusses the impacts of head posture, illumination, occlusion, blur, and individual bias on the accuracy of eye-tracking. The prospective development on eye-tracking employing unsupervised learning, self-supervised learning, and meta-learning have also been discussed.