DESEMPENHO DA TÉCNICA DEEP LEARNING NA ANÁLISE E CATEGORIZAÇÃO DE IMAGENS DE DEFEITO DE MADEIRA

Abstract

Artificial intelligence has made great strides in its field of research, contributing greatly in several areas, such as the analysis and categorization of digital images using machine learning. There are several specific techniques for image recognition and their categorization, one of these techniques, which uses artificial neural networks, involves the specific study, the extraction of characteristics through the analysis of image data of the object being analyzed and the specification of what will be the impact of these characteristics on the neural model for each of the categories, which requires the immersion of the researcher in an area or field of research that is not in his domain. Deep Learning using artificial convolutional neural networks has the ability to learn and extract features during training, without specifying these features in the model, and they generally present better results than those observed by neural network models which had the features observed and programmed by humans. The objective of this work is to apply Deep Learning with the help of the Python language and two libraries called Keras and NumPy, in the categorization of a set of images of wood boards, evaluating its performance. Some convolutional neural network models were developed and evaluated in this process, obtaining promising results, where the best of them presented a categorization error in the order of five percent.