Precision Gaze Localization for Gaze Tracking

This study explores the impact of camera optics on neural network-based gaze-tracking systems. It focuses onthe relationship between visual angle and focal length acrossdifferent camera types. The study analyzes the effects of depthof field (DOF), aperture settings, and optical aberrations such as chromatic and spherical aberration on gaze tracking accuracy. The goal is to standardize optimal camera settings for accurate gaze tracking by establishing the relationship between different camera optics and visual angles. The advanced gaze tracking system, developed using Convolutional Neural Networks (CNNs), achieves an R-squared value of 0.9363, a Mean Squared Error (MSE) of 0.0079, and a Root Mean Squared Error (RMSE) of 0.0890. This system not only enhances the accuracy of gaze estimation but also offers valuable insights into optimizing camera-based eye-tracking technology across various lighting and optical scenarios, paving the way for standardized camera setups in gaze-tracking applications.