Abstract: One of the critical problems prevailing in India is the deaths caused by road accidents. Almost 80% of the accidents are caused by the inattentiveness of the driver. Usage of mobile phones, talking to passengers, reaching behind to grab something and drinking while driving are some of the reasons due to which driver may lose attention. Distractions are of numerous types, out of which we focus on the manual distraction which is based on the posture of the driver. In this paper, we propose a system where we make use of Convolutional Neural Networks and data augmentation techniques. Data augmentation techniques are used to increase the variability of the dataset and decrease overfitting. We have used the first publicly available dataset as input for our model. Our aim is to categorize a test image into one of the nine distinct distracted states of the driver that we have considered. Conclusively, the experimental analysis has shown that applying data augmentation techniques, the proposed model gives better results.
Keywords: Convolutional neural networks, data augmentation techniques, deep learning methods, distracted driver.