Deep Learning para classificação e processamento de imagens
Carregando...
Arquivos
Data
2022-06-20
Autores
Braga, Bernardo Mendonça
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Uma clínica de diagnóstico por imagem em odontologia identifica seus pacientes através de imagens. Essas imagens precisam ser processadas e ordenadas para serem então associadas aos exames de cada paciente, o que requer trabalho manual e repetitivo. O objetivo deste trabalho é a criação de uma ferramenta de classificação de imagens para auxiliar na organização dessas imagens. Para realizar esta classificação redes neurais convolucionais serão utilizadas, pois são capazes de interpretar imagens como dados de entrada. Também será analisado o algoritmo k-means para fins de comparação de técnicas de classificação. Este trabalho visa transformar uma atividade manual repetitiva e de longa duração em um algoritmo. O método de aprendizagem supervisionada utilizando redes neurais convolucionais foi avaliado como o mais adequado para realizar a tarefa proposta. Deep Learning e Machine Learning são esferas dentro da grande área da Inteligência Artificial e outro objetivo deste trabalho é o aprofundamento do estudo dessa área e suas aplicações práticas utilizando a linguagem de programação Python e suas bibliotecas Tensorflow e Keras.
A private dental imaging service provider uses images to identify their patients. Such images need to be processed and classified before they are associated to individual patients’ files, and this has proven to be a repetitive, unqualified job. The focus of the present project is the development of an image classifying tool that may help organize such images. The classification tool will resort to the use of convolutional neural networks as these are capable of interpreting images as entry data. The k-means algorithm will also be analyzed for better comparison between both classification strategies. The objective with this work is, therefore, to turn a repetitive and painstaking job into an algorithm. The supervised learning method with convolutional neural networks was chosen as the more adequate technique to solve the problem presented. Deep learning and machine learning are both part of the artificial intelligence field and this work aims to deepen the understanding of such field and its practical applications through the use of Python programming language and its libraries Tensorflow and Keras.
A private dental imaging service provider uses images to identify their patients. Such images need to be processed and classified before they are associated to individual patients’ files, and this has proven to be a repetitive, unqualified job. The focus of the present project is the development of an image classifying tool that may help organize such images. The classification tool will resort to the use of convolutional neural networks as these are capable of interpreting images as entry data. The k-means algorithm will also be analyzed for better comparison between both classification strategies. The objective with this work is, therefore, to turn a repetitive and painstaking job into an algorithm. The supervised learning method with convolutional neural networks was chosen as the more adequate technique to solve the problem presented. Deep learning and machine learning are both part of the artificial intelligence field and this work aims to deepen the understanding of such field and its practical applications through the use of Python programming language and its libraries Tensorflow and Keras.
Descrição
Palavras-chave
Classificação de Imagens, Inteligência Artificial, Deep Learning, Redes neurais convolucionais, Image classification