Aplicação de redes neurais artificiais profundas na detecção de placas de pare

Abstract

Intelligent systems are increasingly present in our daily lives. Among many, stands out the ones based on artificial neural networks. In this way, we apply artificial intelligence techniques in the development of a computer vision solution for object detection in images, with focus on stop signs. Artificial neural network was used to develop a classifier capable of integrating features extracted from images, performing the classification and localization of objects in them. A multilayered convolutional artificial neural network with deep learning techniques was used. The projected neural network follows procedure set in the YOLO (You Only Look Once) project, taking advantage of its fast learning capabilities and extensive image pattern recognition, achieving a good success rate. Experiments allowed detection of stop signs invariably to space in complex images.