Research on deep learning models and data preprocessing techniques for driver fatigue detection

Detecting signs of fatigue while driving is one of the most important and urgent issues in today's modern transportation system. Early detection and warning will significantly reduce the risk of traffic accidents. To accomplish this, several methods have been proposed. In recent years, methods based on deep learning techniques have attracted great attention due to their high efficiency and cost savings in detecting and warning about signs of fatigue while driving. In this paper, we study a method to improve the accuracy of deep learning models based on the data preprocessing technique, applied to detect signs of fatigue while driving. Data preprocessing helps the training images become more accurate, noise-reduced, diverse and rich instead of just using the main image. To ensure high reliability and accuracy, the methods are trained and tested on large datasets including thousands of images of all types. Experimental results show that the combination of the VGG-16 model and the data augmentation technique gives better results than the traditional deep learning model, with an accuracy of over 96%.