Driver Drowsiness Detection System Using Deep Learning Method to Reduce Risk Accident

Fatigue is a common problem that often causes drowsiness in individuals. This condition can have serious repercussions, especially in activities that require high alertness such as driving. In research, we propose a system that detects the condition of the driver on a means of transportation. The system uses You Only Look Once Version Five (YOLOv5) algorithm that has been turned off with thousands of images from the dataset of open eyes, closed eyes, and yawning mouth with different hyperparameters for performance to be analyzed in order to get the most optimal model. As a result, the YOLOv5 deep learning model produces an accuracy rate of 0.737 which will detect the driver’s condition and when sleepiness is detected, the buzzer will light up as a reminder and the time of the incident is sent to the database using Long Range (LoRa) communication.