Utilização de Técnicas de Data Augmentation em Imagens: Teoria e Prática

Abstract

First, I would like to thank my family: my grandmother Rosalha Borges, my mother Luciane Borges, my aunt Crisitiane Borges, and my brother Diogo Borges, for always being there for me and supporting me in life, especially in difficult moments such as this one where I had to absent from their presence to accomplish this work. Without them I would not be able to fulfill this work. I would like to thank my adviser, Prof. Markus Endler, to whom I have an immense gratitude and admiration for his excellent guidance, availability, patience and for having given me the honor of working with him. His research vision has inspired and shaped me into becoming a better researcher. I'm sure our research partnership will continue after the end of my PhD. I would also like to thank Prof. Francisco Silva. His critical view guided and shaped numerous parts of this thesis. I learned a lot from him. In the same way, I would also like to thank Prof. Marco Casanova and Prof. Hélio Lopes for encouraging me to work with this topic and providing numerous suggestions, especially the insight for the density heuristic. I would like to thank my roommates during the PhD journey: Àlan Guedes, Eduardo Araújo, Derlyane Simão, Katia Vega, and Thais Abreu. I learned a lot living with them and they made the PhD journey with more joy and happiness. I will miss you all. In addition to my funny roommates, I would also like to thank all my friends (1801!

Um Estudo de Caso para Detecção de Ovócitos de Peixe Utilizando Redes Neurais Convolucionais

“…Um dos atributos presentes nos rótulos representa as fases de desenvolvimento maturacional ovocitário. As fases utilizadas nesta pesquisa foram escolhidas devido a frequência destes ovócitos nas imagens, pois segundo Claro et al [Claro et al 2020] técnicas baseadas em aprendizagem profunda necessitam de uma grande quantidade de objetos para uma detecc ¸ão mais eficiente.…”