Vision-Based Drowsiness Detection System Using Convolutional Neural Networks

Abstract

Drowsiness and sleeping cause traffic accidents that result in the deaths and injuries of people. Nowadays, many different systems are being developed for intelligent cars to help drivers to reduce traffic accidents. This paper represents real-time drowsiness detection based on computer vision. This system is used to detect drowsiness with a camera which is located in front of the driver and alerts the driver if there is an act of drowsiness. The system detects drowsiness by means of face and eyes from the specific area of the image and eye blinks of the driver. Viola and Jones face detector [1] has been implemented to detect the faces, then Convolutional Neural Network (CNN) is used for detection of drowsiness. The system was tested on different conditions such as a driver with glasses, a driver without glasses, also on different lighting conditions. This system is a robust, automatic, and can operate without any calibration.