REAL-TIME DETECTION OF DRIVER DISTRACTION USING RESNET 50-BASED DEEP LEARNING MODEL

Road-related deaths are on the rise largely due to people being distracted while driving. People are also increasingly using mobile devices and infotainment systems in their vehicles. Thus, there is a growing need for systems which can monitor driver behavior in real-time and help improve their safety. This thesis demonstrates the application of deep learning for detecting signs that a driver may be distracted. Using a ResNet50 convolutional neural network (CNN) as the foundation of our deep learning framework, we created a real-time method of detecting visual signs of driver distraction via the use of photographs taken with cameras installed in vehicles. The Res Net 50 model utilizes a residual learning technique and transfer learning methodology. It has been shown that the ResNet50 model has outperformed other forms of machine learning and fewer layers of CNNs when completing tasks related to detecting driver distraction. Through additional findings from previous studies, this research provides a basis for future work to develop advanced driver assistance systems (ADAS) to improve road safety through increasing our knowledge about how to monitor driver attention levels.