Detecção de câncer de pele usando Cluster Raspberry PI e a plataforma Pytorch

• Ano de defesa: 2022
• Autor(a) principal: Matos, Elias Rabelo
• Orientador(a): Ordonez, Edward David Moreno
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Programa de Pós-Graduação: Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: Computação de alto desempenho; Inteligência artificial; Câncer de pele; Raspberry PI (computador)
• Palavras-chave em Inglês: Artifical intelligence; High-performance computing; Skin Cancer
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://ri.ufs.br/jspui/handle/riufs/22448

Skin cancer causes deaths and has an increase in the number of cases each year worldwide. One of the most common ways to detect skin cancer is the use of the ABCDE rule, applied to a set of features on a patient’s skin lesion. From this form of diagnosis, Skin Cancer can be treat by Computer Science as an image classification problem using Artificial Intelligence. Since 2016 the ISIC (International Skin Cancer Challenge) has launched annual competitions towards the detection of skin cancer. In this work, the selected dataset is the HAM1000, which contains skin images divided into seven different types of lesions and which is part of the ISIC files. With this dataset, the objectives of the work were defined. Training of a convolutional neural network in a Low-Cost High-Performance Cluster using the Raspberry Pi platform; and apply the knowledge transfer technique from a convolutional neural network to an MLP network running on the Raspberry PI platform. Cluster training was not possible with the configuration used in this job. And with the transfer of knowledge, was achieved the value of 80% accuracy in the best settings used.