Publication

Abstract : In India, every year approximately 1.35 million people lose their lives due to road accidents. Many accidents are due to drunk and drive cases, overspeed cases, and negligence of traffic rules. Most of the fatal accidents are due to drowsiness of the driver. The drowsiness can be due to the consumption of alcohol or may be due to fatigue. A smart accident prevention system can be developed if the drowsiness of the driver is detected. This paper presents machine learning-based driver drowsiness detection. Convolutional Neural Network (CNN) and Haar Cascade Classifier are used to detect the drowsiness of the driver. The real-time images are captured. Images of different stages of eyes such as opened eyes, and closed eyes at normal and dark conditions are captured and used to train the classifier model. Blurred images are also taken to train the model. The original data is augmented to create a larger dataset for training the models. Therefore, the dataset created is a good mix of all possible conditions. CNN classifier performance is found to be better than the Haar cascade classifier to detect the drowsiness of the driver effectively. Haar Classifier cannot classify the blurry images or the images that are taken during the night.