Classification of Road Incidents Using Deep Learning Classifiers Compared to SVM

Abstract

Accidents on the roads affect the lives of almost 1.3 million people annually, as reported by World Health Organization (WHO) preliminary data, and minimizing vehicle crashes is still a significant challenge due to the complex interaction of many factors, among these reasons is the delayed response of emergency services. Thus, there is a need to improve the accuracy of detection systems, in this contribution we propose classifiers with high precision that can be integrated into real time accident detection systems. We create three different models: A Support Vector Machine SVM, a Convolutional Neural Network CNN, and a Recurrent Convolution Neural Network R-CNN. Our goals are to highlight the two insights deep learning algorithms and traditional classification techniques, and to improve road accident detection accuracy by merging these models using the advantages of ensemble learning.