Detection of Driver Distraction Using Convolutional Neural Network

Abstract

Driver distraction is a major factor in vehicle accidents. Day by day the percentage of vehicle crashes is rising due to several activities performed by the driver. Many researchers have developed various algorithms to detect driver distraction. In this paper, an algorithm is proposed to detect driver distraction in real time. To evaluate the proposed algorithm, the dataset is used which was provided by State Farm through a Kaggle competition. The predefined convolutional neural network (CNN) namely resnet50 is used to extract the features of testing images and training images. The extracted features are fed to a Support vector machine (SVM) to classify the normal images and distracted images. The proposed approach classification accuracy achieved 84% for normal images and 87% for distracted images. A real time hardware model is developed by using Arduino Uno to detect the driver distraction. The output hardware modules such as LED, LCD display, and buzzer are used to alert the driver.