Comparando algoritmos de aprendizado profundo para o problema de detecção de distração de motoristas a partir de imagens

Abstract

One of the major causes of traffic accidents today is the distraction of drivers while driving. It may be the distraction by using cell phones behind the wheel or by talking looking at the passenger. One possible way to reduce the number of accidents is to alert the driver when he is distracted. The most difficult task of telling the driver when he is distracted is to detect distraction. One way to achieve this is through the use of Deep Learning. Thus, this work aims to discover the best technique for classification of images for the problem of detecting the distraction of drivers. To this end, several architectures of Convolutional Neural Networks were tested to find out the best architecture to solve such problem.