Estudo comparativo entre redes neurais convolucionais para um problema de classificação

• Autor(a) principal: Lima, Francisco Igor da Silva
• Data de Publicação: 2018
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal do Ceará (UFC)
• Texto Completo: http://www.repositorio.ufc.br/handle/riufc/39460
• topic: Aprendizagem Profunda; Redes Neurais (Computação); Visão Computacional

The use of convolutional neural networks for image classification has grown up in the last years. Since 2012 there are notable great advances in the computational vision field. From this period, convolutional neural networks have gained the attention of the academy and the market. Due, principally, to its good results, new algorithms, and greater computational power, the last one, especially caused by the advantages of cloud computing. This work aims to make a comparative study among the neural networks: VGG16, InceptionV3, ResNet50 to discover what is the best configuration among these networks to a solution of an image classification problem. As result it is possible to affirm that VGG16 is not a good network for this problem, while InceptionV3 and ResNet50 present better results.