E-Blink details

Eye Tracking System For Monitoring Eye Fatigue Using Image Processing Methods

The high use of smartphones in Indonesia brings risks to Visual Display Terminal (VDT) users, such as Computer Vision Syndrome (CVS) which causes headaches, dry eyes, and eye strain. To reduce CVS symptoms, it is recommended to optimize the work environment, screen position, blink frequently, take breaks, and use artificial eye drops. A decrease in blinking frequency when using a Visual Display Terminal (VDT) is a key indicator of CVS. The purpose of this study is to design and implement a Convolutional Neural Network (CNN) model integrated on the Raspberry Pi 3 Model B+ platform to detect eye fatigue in VDT users.

The experimental method was conducted by collecting blink frequency data through a camera connected to the Raspberry Pi 3. This model has a validation accuracy of up to 99.26%. In several experiments, the system on the Raspberry Pi was able to predict and calculate the frequency of eye blinks by 75.1%, and the average latency was 0.19 seconds per frame. Despite being optimized with post-training quantization, the prediction accuracy decreased especially in low light conditions. The implementation of this tool is designed to provide early warning to users, reducing the negative impact of overexposure to digital screens