Pendeteksian Mata dan Wajah Secara Kontinu untuk Pemantauan Penggunaan Komputer dengan Menggunakan Algoritma Mobilenet

Abstract

Computer Vision Syndrome is becoming an increasingly alarming issue due to prolonged screen exposure, leading to eye strain, headaches, and visual discomfort. This system is designed with real-time detection capabilities to monitor the user's eye and facial positioning, issuing alerts in accordance with the implemented 20-20-20 rule. The MobileNet algorithm was chosen for its high efficiency, making it well-suited for deployment on resource-constrained devices, enabling precise detection without compromising system performance. MobileNet is employed in the development of two detection models: the Eye Gaze model and the Eye Blink model. The Eye Gaze model is designed to accurately determine the direction of a user's gaze, while the Eye Blink model detects blinking frequency and eye conditions, distinguishing between open and closed states. System testing yielded an accuracy rate of 100% for the Eye Gaze model based on 831 test cases, and 99.07% accuracy for the Eye Blink model based on 218 test cases. Based on these results, the system can be deemed highly optimal in detecting gaze direction and blinking as a supportive measure for the 20-20-20 method in the prevention of Computer Vision Syndrome.