Semantic Segmentation based on Variational Methods

• Grant number: 19/08589-0
• Support Opportunities: Scholarships in Brazil - Master
• Start date: September 01, 2019
• End date: November 30, 2021
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Gerberth Adín Ramírez Rivera
• Grantee: Rodrigo Fumihiro de Azevedo Kanehisa
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Semantic segmentation is one of the great open-ended problems in computer vision. This problem consists in classifying each pixel present in an image, delimiting an object. Convolutional neural networks have been shown to be efficient in detecting, identifying and segmenting objects in images. However, during the segmentation process, some problems arise, such as loss of spatial information and low resolution of feature maps. These problems lead to rough or noisy segmentations due to the uncertainties on the data. One way to counter these problems is to use statistical methods in conjunction with neural networks. This project proposes neural networks based on variational methods for semantic segmentation. Variational methods present a solution to map the contents of the image into a latent space that represents distributions of the data. Thus, allowing the network to handle uncertainty and model more complex information. We propose to use a Mixture of Gaussian Models to tackle the complex information. We will evaluate our proposals within existing databases and compare against existing methods on standard benchmarks for semantic segmentation.

Academic Publications

(References retrieved automatically from State of São Paulo Research Institutions)

KANEHISA, Rodrigo Fumihiro de Azevedo. Segmentação semantica com mixtura global de priors Gaussianas. 2022. Master's Dissertation - Universidade Estadual de Campinas (UNICAMP). Instituto de Computação Campinas, SP.