Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation

Semantic segmentation performs pixel-wise classification for given images, which can be widely used in autonomous driving, robotics, medical diagnostics and etc. The recent advanced approaches have witnessed rapid progress in semantic segmentation. However, these supervised learning based methods re...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Zhiying Cao, Tengfei Zhang, Wenhui Diao, Yue Zhang, Xiaode Lyu, Kun Fu, Xian Sun
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2019
Materias:	Meta-learning few-shot semantic segmentation Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/c733ea5763c84041ba4ff1446090b10c
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:c733ea5763c84041ba4ff1446090b10c
record_format	dspace
spelling	oai:doaj.org-article:c733ea5763c84041ba4ff1446090b10c2021-11-19T00:02:49ZMeta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation2169-353610.1109/ACCESS.2019.2953465https://doaj.org/article/c733ea5763c84041ba4ff1446090b10c2019-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/8901116/https://doaj.org/toc/2169-3536Semantic segmentation performs pixel-wise classification for given images, which can be widely used in autonomous driving, robotics, medical diagnostics and etc. The recent advanced approaches have witnessed rapid progress in semantic segmentation. However, these supervised learning based methods rely heavily on large-scale datasets to acquire strong generalizing ability, such that they are coupled with some constraints. Firstly, human annotation of pixel-level segmentation masks is laborious and time-consuming, which causes relatively expensive training data and make it hard to deal with urgent tasks in dynamic environment. Secondly, the outstanding performance of the above data-hungry methods will decrease with few available training examples. In order to overcome the limitations of the supervised learning semantic segmentation methods, this paper proposes a generalized meta-learning framework, named Meta-Seg. It consists of a meta-learner and a base-learner. Specifically, the meta-learner learns a good initialization and a parameter update strategy from a distribution of few-shot semantic segmentation tasks. The base-learner can be any semantic segmentation models theoretically and can implement fast adaptation (that is updating parameters with few iterations) under the guidance of the meta-learner. In this work, the successful semantic segmentation model FCN8s is integrated into Meta-Seg. Experiments on the famous few-shot semantic segmentation dataset PASCAL<inline-formula> <tex-math notation="LaTeX">$5^{i}$ </tex-math></inline-formula> prove Meta-Seg is a promising framework for few-shot semantic segmentation. Besides, this method can provide with reference for the relevant researches of meta-learning semantic segmentation.Zhiying CaoTengfei ZhangWenhui DiaoYue ZhangXiaode LyuKun FuXian SunIEEEarticleMeta-learningfew-shotsemantic segmentationElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 7, Pp 166109-166121 (2019)
institution	DOAJ
collection	DOAJ
language	EN
topic	Meta-learning few-shot semantic segmentation Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Meta-learning few-shot semantic segmentation Electrical engineering. Electronics. Nuclear engineering TK1-9971 Zhiying Cao Tengfei Zhang Wenhui Diao Yue Zhang Xiaode Lyu Kun Fu Xian Sun Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation
description	Semantic segmentation performs pixel-wise classification for given images, which can be widely used in autonomous driving, robotics, medical diagnostics and etc. The recent advanced approaches have witnessed rapid progress in semantic segmentation. However, these supervised learning based methods rely heavily on large-scale datasets to acquire strong generalizing ability, such that they are coupled with some constraints. Firstly, human annotation of pixel-level segmentation masks is laborious and time-consuming, which causes relatively expensive training data and make it hard to deal with urgent tasks in dynamic environment. Secondly, the outstanding performance of the above data-hungry methods will decrease with few available training examples. In order to overcome the limitations of the supervised learning semantic segmentation methods, this paper proposes a generalized meta-learning framework, named Meta-Seg. It consists of a meta-learner and a base-learner. Specifically, the meta-learner learns a good initialization and a parameter update strategy from a distribution of few-shot semantic segmentation tasks. The base-learner can be any semantic segmentation models theoretically and can implement fast adaptation (that is updating parameters with few iterations) under the guidance of the meta-learner. In this work, the successful semantic segmentation model FCN8s is integrated into Meta-Seg. Experiments on the famous few-shot semantic segmentation dataset PASCAL<inline-formula> <tex-math notation="LaTeX">$5^{i}$ </tex-math></inline-formula> prove Meta-Seg is a promising framework for few-shot semantic segmentation. Besides, this method can provide with reference for the relevant researches of meta-learning semantic segmentation.
format	article
author	Zhiying Cao Tengfei Zhang Wenhui Diao Yue Zhang Xiaode Lyu Kun Fu Xian Sun
author_facet	Zhiying Cao Tengfei Zhang Wenhui Diao Yue Zhang Xiaode Lyu Kun Fu Xian Sun
author_sort	Zhiying Cao
title	Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation
title_short	Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation
title_full	Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation
title_fullStr	Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation
title_full_unstemmed	Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation
title_sort	meta-seg: a generalized meta-learning framework for multi-class few-shot semantic segmentation
publisher	IEEE
publishDate	2019
url	https://doaj.org/article/c733ea5763c84041ba4ff1446090b10c
work_keys_str_mv	AT zhiyingcao metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation AT tengfeizhang metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation AT wenhuidiao metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation AT yuezhang metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation AT xiaodelyu metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation AT kunfu metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation AT xiansun metasegageneralizedmetalearningframeworkformulticlassfewshotsemanticsegmentation
_version_	1718420693757984768

Meta-Seg: A Generalized Meta-Learning Framework for Multi-Class Few-Shot Semantic Segmentation

Ejemplares similares