Meta learning approaches for few-shot semantic segmentation with sparse labels
Autor(a) principal: | |
---|---|
Data de Publicação: | 2021 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UFMG |
Texto Completo: | http://hdl.handle.net/1843/58285 https://orcid.org/0000-0002-9802-593X |
Resumo: | Semantic Segmentation is a classic task in Computer Vision that has multiple applications in many areas, from organ segmentation for clinical image studies, or counting objects in production lines, to estimating deforestation areas sizes. However, the type of data labeling required for actual methods to solve this problem is laborious to produce, since one has to determine the label for all pixels of an image. This usually increases the cost (human and/or monetary) to produce new datasets. Two possible ways to reduce this cost are: 1) reducing the number of labeled samples; 2) using simpler/sparse types of annotation. Despite that, current and usual deep learning based methods for segmentation tend to perform poorly when using one, or two, of these solutions. In this work, we propose two meta learning methods to the few-shot semantic segmentation task with sparse annotations. These two approaches are based on two existing methods for classification: Model-Agnostic Meta-Learning (MAML) and Prototypical Networks. Our methods were tested in different scenarios in the medical and remote sensing areas, which usually have limited data access, and obtained competitive results in different tasks. |
id |
UFMG_2bd46bf3c1c86d14950c63cbb35d70a4 |
---|---|
oai_identifier_str |
oai:repositorio.ufmg.br:1843/58285 |
network_acronym_str |
UFMG |
network_name_str |
Repositório Institucional da UFMG |
repository_id_str |
|
spelling |
Jefersson Alex dos Santoshttp://lattes.cnpq.br/2171782600728348Hugo Neves de OliveiraAdriano Alonso VelosoGilson Alexandre Ostwald Pedro Costahttp://lattes.cnpq.br/8298724730331584Pedro Henrique Targino Gama2023-08-28T14:28:24Z2023-08-28T14:28:24Z2021-04-15http://hdl.handle.net/1843/58285https://orcid.org/0000-0002-9802-593XSemantic Segmentation is a classic task in Computer Vision that has multiple applications in many areas, from organ segmentation for clinical image studies, or counting objects in production lines, to estimating deforestation areas sizes. However, the type of data labeling required for actual methods to solve this problem is laborious to produce, since one has to determine the label for all pixels of an image. This usually increases the cost (human and/or monetary) to produce new datasets. Two possible ways to reduce this cost are: 1) reducing the number of labeled samples; 2) using simpler/sparse types of annotation. Despite that, current and usual deep learning based methods for segmentation tend to perform poorly when using one, or two, of these solutions. In this work, we propose two meta learning methods to the few-shot semantic segmentation task with sparse annotations. These two approaches are based on two existing methods for classification: Model-Agnostic Meta-Learning (MAML) and Prototypical Networks. Our methods were tested in different scenarios in the medical and remote sensing areas, which usually have limited data access, and obtained competitive results in different tasks.Segmentação Semântica é uma tarefa clássica de visão computacional que tem múltiplas aplicações em diversas áreas, desde de segmentação de órgãos para estudos clínicos por imagem, contagem de objetos em linha de produção, até a estimativa de tamanho de áreas de desmatamento. Entretanto, o tipo de rotulação de dados necessária para os métodos atuais resolverem o problema é laboriosa de se produzir, uma vez que é necessário determinar os rótulos de todos os pixels da imagem. Isso costuma aumentar o custo (humano e/ou monetário) de construção de novos conjuntos de dados. Duas formas possíveis de se reduzir esse custo são: 1) diminuindo o número de imagens anotadas; 2) usando um formato de anotação mais simples/esparsa. Porém, os métodos comuns e mais atuais, de deep learning, para segmentação semântica não funcionam bem usando uma, ou duas, dessas soluções. Neste trabalho propomos dois métodos de meta learning para segmentação semântica em cenários few-shot com rotulação esparsa. Essas abordagens foram baseadas em dois métodos existentes para classificação: Model-Agnostic Meta-Learning (MAML) e Prototypical Networks. As nossas abordagens foram testadas em diversos cenários da área médica e sensoriamento remoto, que normalmente tem uma limitação de aquisição de dados, e obtiveram resultados competitivos em diferentes tarefas.CNPq - Conselho Nacional de Desenvolvimento Científico e TecnológicoCAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível SuperiorengUniversidade Federal de Minas GeraisPrograma de Pós-Graduação em Ciência da ComputaçãoUFMGBrasilICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃOComputação – TesesVisão por computador – TesesAprendizagem de máquina – TesesAprendizado profundo – TesesMeta-aprendizado – TesesComputingComputer VisionMachine LearningDeep LearningMeta LearningMeta learning approaches for few-shot semantic segmentation with sparse labelsinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisinfo:eu-repo/semantics/openAccessreponame:Repositório Institucional da UFMGinstname:Universidade Federal de Minas Gerais (UFMG)instacron:UFMGORIGINALdissertação_mestrado_final.pdfdissertação_mestrado_final.pdfapplication/pdf33612694https://repositorio.ufmg.br/bitstream/1843/58285/1/disserta%c3%a7%c3%a3o_mestrado_final.pdf3cb388f98d155a274c06dc6d22a0249bMD51LICENSElicense.txtlicense.txttext/plain; charset=utf-82118https://repositorio.ufmg.br/bitstream/1843/58285/2/license.txtcda590c95a0b51b4d15f60c9642ca272MD521843/582852023-08-28 11:28:25.208oai:repositorio.ufmg.br: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ório de PublicaçõesPUBhttps://repositorio.ufmg.br/oaiopendoar:2023-08-28T14:28:25Repositório Institucional da UFMG - Universidade Federal de Minas Gerais (UFMG)false |
dc.title.pt_BR.fl_str_mv |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
title |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
spellingShingle |
Meta learning approaches for few-shot semantic segmentation with sparse labels Pedro Henrique Targino Gama Computing Computer Vision Machine Learning Deep Learning Meta Learning Computação – Teses Visão por computador – Teses Aprendizagem de máquina – Teses Aprendizado profundo – Teses Meta-aprendizado – Teses |
title_short |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
title_full |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
title_fullStr |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
title_full_unstemmed |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
title_sort |
Meta learning approaches for few-shot semantic segmentation with sparse labels |
author |
Pedro Henrique Targino Gama |
author_facet |
Pedro Henrique Targino Gama |
author_role |
author |
dc.contributor.advisor1.fl_str_mv |
Jefersson Alex dos Santos |
dc.contributor.advisor1Lattes.fl_str_mv |
http://lattes.cnpq.br/2171782600728348 |
dc.contributor.advisor-co1.fl_str_mv |
Hugo Neves de Oliveira |
dc.contributor.referee1.fl_str_mv |
Adriano Alonso Veloso |
dc.contributor.referee2.fl_str_mv |
Gilson Alexandre Ostwald Pedro Costa |
dc.contributor.authorLattes.fl_str_mv |
http://lattes.cnpq.br/8298724730331584 |
dc.contributor.author.fl_str_mv |
Pedro Henrique Targino Gama |
contributor_str_mv |
Jefersson Alex dos Santos Hugo Neves de Oliveira Adriano Alonso Veloso Gilson Alexandre Ostwald Pedro Costa |
dc.subject.por.fl_str_mv |
Computing Computer Vision Machine Learning Deep Learning Meta Learning |
topic |
Computing Computer Vision Machine Learning Deep Learning Meta Learning Computação – Teses Visão por computador – Teses Aprendizagem de máquina – Teses Aprendizado profundo – Teses Meta-aprendizado – Teses |
dc.subject.other.pt_BR.fl_str_mv |
Computação – Teses Visão por computador – Teses Aprendizagem de máquina – Teses Aprendizado profundo – Teses Meta-aprendizado – Teses |
description |
Semantic Segmentation is a classic task in Computer Vision that has multiple applications in many areas, from organ segmentation for clinical image studies, or counting objects in production lines, to estimating deforestation areas sizes. However, the type of data labeling required for actual methods to solve this problem is laborious to produce, since one has to determine the label for all pixels of an image. This usually increases the cost (human and/or monetary) to produce new datasets. Two possible ways to reduce this cost are: 1) reducing the number of labeled samples; 2) using simpler/sparse types of annotation. Despite that, current and usual deep learning based methods for segmentation tend to perform poorly when using one, or two, of these solutions. In this work, we propose two meta learning methods to the few-shot semantic segmentation task with sparse annotations. These two approaches are based on two existing methods for classification: Model-Agnostic Meta-Learning (MAML) and Prototypical Networks. Our methods were tested in different scenarios in the medical and remote sensing areas, which usually have limited data access, and obtained competitive results in different tasks. |
publishDate |
2021 |
dc.date.issued.fl_str_mv |
2021-04-15 |
dc.date.accessioned.fl_str_mv |
2023-08-28T14:28:24Z |
dc.date.available.fl_str_mv |
2023-08-28T14:28:24Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/1843/58285 |
dc.identifier.orcid.pt_BR.fl_str_mv |
https://orcid.org/0000-0002-9802-593X |
url |
http://hdl.handle.net/1843/58285 https://orcid.org/0000-0002-9802-593X |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Minas Gerais |
dc.publisher.program.fl_str_mv |
Programa de Pós-Graduação em Ciência da Computação |
dc.publisher.initials.fl_str_mv |
UFMG |
dc.publisher.country.fl_str_mv |
Brasil |
dc.publisher.department.fl_str_mv |
ICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃO |
publisher.none.fl_str_mv |
Universidade Federal de Minas Gerais |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFMG instname:Universidade Federal de Minas Gerais (UFMG) instacron:UFMG |
instname_str |
Universidade Federal de Minas Gerais (UFMG) |
instacron_str |
UFMG |
institution |
UFMG |
reponame_str |
Repositório Institucional da UFMG |
collection |
Repositório Institucional da UFMG |
bitstream.url.fl_str_mv |
https://repositorio.ufmg.br/bitstream/1843/58285/1/disserta%c3%a7%c3%a3o_mestrado_final.pdf https://repositorio.ufmg.br/bitstream/1843/58285/2/license.txt |
bitstream.checksum.fl_str_mv |
3cb388f98d155a274c06dc6d22a0249b cda590c95a0b51b4d15f60c9642ca272 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFMG - Universidade Federal de Minas Gerais (UFMG) |
repository.mail.fl_str_mv |
|
_version_ |
1803589439747784704 |