Classificação de placas de trânsito com redes neurais para automação de veículos

Abstract

This article aims to develop a study on the functioning of Convolutional Neural Networks in the classification of traffic signs for use in autonomous cars, thus being able to be used as a tool to aid the vehicle in the decision process while driving, either through controlling the vehicle's speed, identifying the changes that the vehicle must make or even displaying information to the passenger through an on-board computer. To carry out this project, the German Traffic Sign dataset [1] was used. It is a public dataset used during the International Joint Conference on Neural Networks (IJCNN) in 2011. It has 39209 images for model training and 12630 images for testing, both divided into 43 distinct classes, each one indicating a different type. of traffic sign. At the end of the article, the reader will not only have knowledge about the theory behind neural networks, but also a theoretical background on the practical use of Convolutional Neural Networks for image classification.