A Real-Time Analytics Approach for Computer Vision Syndrome Detection

The widespread use of digital devices has increased the prevalence of Computer Vision Syndrome (CVS), characterized by symptoms such as low blink rates, eye strain, and irritation. Existing CVS prevention systems often focus solely on blink rate, limiting the range of symptoms detected. This paper presents a real-time CVS identification system that integrates multiple visual metrics, including pupil dilation, red-eye detection, and blink rate, for a more accurate and holistic CVS detection. A method based on eye and iris landmark detection relies on a pre-trained convolutional neural network model to track the eye region. The proposed method was demonstrated by training data from five racial groups using fifty equally distributed images between male and female subjects. The fastest detection was performed for Caucasian males, and the slowest was observed on East Asian male subjects, with the detection time consistently under one second. In detecting pupil dilation, the accuracy of the proposed method is 78% based on videos of normal and dilated pupils. Designed for efficiency with training data across diverse demographics, the system achieved an overall performance score of 84%, achieving strong detection precision and wide coverage of symptoms.