Classification of Images to Detect Distracted Drivers

Vehicle accidents are a major concern in current transportation systems. The World Health Organization reports that road accidents are the eighth leading cause of death globally. Many of these accidents, over 80%, are caused by distracted driving such as using a mobile phone, talking to passengers, or smoking while behind the wheel. While efforts have been made to address this issue, there is no perfect solution. One potential approach is to use quantitative measures to assess driver activities and create a classification system that can detect distracting actions. In this project, a range of deep learning models are implemented and tested with various different parameter values, that can effectively classify driver distractions and increase driver awareness for improved safety. As a result of this experiment, it has been observed that base Convolutional Neural Network (CNN) model with seven convolutional layers when trained using augmented data performed the best in terms of accuracy. However, ResNet50 outperformed VGG16 based model when comparing transfer learning based models.