Gaze Tracking Using Capsule-Based CNN Architecture for HMI

Gaze tracking is one of the most widely used method for human–machine interface (HMI) applications in varied areas, such as customer behavioral analysis, driver fatigue analysis, and many more. Gaze tracking provides information about the direction, where a person is looking at with the help of videos or images of the face of the person or only their eyes. A capsule network-based architecture has been developed in this work to predict human gaze direction. We have analyzed the developed architecture by investigating the performance for subject-independent as well subject-dependent scenarios.