Eye Movement analysis And Prediction Using Image Processing
V.K.G. Kalaiselvi, D. Doss · 2024
The predominant function of most facial analysis systems revolves around facial alignment and eye tracking, crucial for locating key facial landmarks in images or videos. While developers have access to various models and techniques within popular computer vision libraries, the real-time demands and challenges posed by extreme head positions, occlusions, and inadequate lighting often limit their effectiveness.In this article, we propose a range of software optimization techniques and training strategies centered on data augmentation. These strategies aim to enhance a diverse array of models integral to real-time face alignment algorithms. We advocate for a broader set of evaluation criteria, enabling innovative assessments that mitigate common issues encountered in real-time monitoring scenarios.Our experimental findings showcase that models developed using our suggested methodologies exhibit smoother tracking, enhanced speed, reduced size, heightened accuracy, and greater resilience in challenging conditions.