Segmentation of images with fully convolutional networks

• Grant number: 20/02891-3
• Support type: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): April 01, 2020
• Effective date (End): December 31, 2020
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Cooperation agreement: Microsoft Research
• Principal researcher: Nina Sumiko Tomita Hirata
• Grantee: Pedro Henrique Barbosa de Almeida
• Home Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Company: Universidade de São Paulo (USP). Instituto de Matemática e Estatística (IME)
• Associated research grant: 17/25835-9 - Understanding images and deep learning models, AP.PITE

Abstract

The goal of this project is to expose the student to a cycle of scientific research that contemplates the study of fundamentals, the development and the application of one or more methods, and evaluation of the results, with comparisons when it is the case. In particular, the focus of the study is on image-to-image transformation problems of which segmentation is an example, and on the application of fully convolutional networks on these problems. We foresee applications in binary segmentation problems such as vessel segmentation in retinal images and text or specific graphic objects in document images. (AU)