Combining clustering and active learning for the detection and learning of new image classes
Autor(a) principal: | |
---|---|
Data de Publicação: | 2019 |
Outros Autores: | , , , |
Tipo de documento: | Artigo |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UNESP |
Texto Completo: | http://dx.doi.org/10.1016/j.neucom.2019.04.070 http://hdl.handle.net/11449/185768 |
Resumo: | Discriminative classification models often assume all classes are available at the training phase. As such models do not have a strategy to learn new concepts from available unlabeled instances, they usually work poorly when unknown classes emerge from future data to be classified. To address the appearance of new classes, some authors have developed approaches to transfer knowledge from known to unknown classes. Our study provides a more flexible approach to learn new (visual) classes that emerge over time. The key idea is materialized by an iterative classifier that combines Support Vector Machines with clustering via an optimization algorithm. An entropy and density-based selection strategy explores label uncertainty and high-density regions from unlabeled data to be classified. Selected instances from new classes are submitted to get labels and then used to improve the model. The proposed image classifier is consistently better than approaches that select instances randomly or from clusters. We also show that features obtained via Deep Learning methods improve results when compared with shallow features, but only using our selection strategy. Our approach requires fewer iterations to learn new classes, thereby significantly saving labeling costs. (C) 2019 Elsevier B.V. All rights reserved. |
id |
UNSP_0d2331b7421d379a9e18c1147d2e13b4 |
---|---|
oai_identifier_str |
oai:repositorio.unesp.br:11449/185768 |
network_acronym_str |
UNSP |
network_name_str |
Repositório Institucional da UNESP |
repository_id_str |
2946 |
spelling |
Combining clustering and active learning for the detection and learning of new image classesImage classificationActive learningClusteringOpen setDeep learningDiscriminative classification models often assume all classes are available at the training phase. As such models do not have a strategy to learn new concepts from available unlabeled instances, they usually work poorly when unknown classes emerge from future data to be classified. To address the appearance of new classes, some authors have developed approaches to transfer knowledge from known to unknown classes. Our study provides a more flexible approach to learn new (visual) classes that emerge over time. The key idea is materialized by an iterative classifier that combines Support Vector Machines with clustering via an optimization algorithm. An entropy and density-based selection strategy explores label uncertainty and high-density regions from unlabeled data to be classified. Selected instances from new classes are submitted to get labels and then used to improve the model. The proposed image classifier is consistently better than approaches that select instances randomly or from clusters. We also show that features obtained via Deep Learning methods improve results when compared with shallow features, but only using our selection strategy. Our approach requires fewer iterations to learn new classes, thereby significantly saving labeling costs. (C) 2019 Elsevier B.V. All rights reserved.Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)Sao Paulo State Univ, Sch Sci & Engn, Tupa, SP, BrazilUniv Sao Paulo, Inst Math & Comp Sci, Sao Carlos, SP, BrazilUniv Sao Paulo, Dept Comp Engn & Digital Syst, Sao Carlos, SP, BrazilUniv Texas Austin, Dept Elect & Comp Engn, IDEAL, Austin, TX 78712 USAUniv Texas Austin, Dept Elect & Comp Engn, Machine Learning Res Grp, Austin, TX 78712 USAUniv Texas Austin, Dept Elect & Comp Engn, Austin, TX 78712 USASao Paulo State Univ, Sch Sci & Engn, Tupa, SP, BrazilFAPESP: 2017/00357-7Elsevier B.V.Universidade Estadual Paulista (Unesp)Universidade de São Paulo (USP)Univ Texas AustinColetta, Luiz F. S. [UNESP]Ponti, MoacirHruschka, Eduardo R.Acharya, AyanGhosh, Joydeep2019-10-04T12:38:21Z2019-10-04T12:38:21Z2019-09-17info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/article150-165http://dx.doi.org/10.1016/j.neucom.2019.04.070Neurocomputing. Amsterdam: Elsevier, v. 358, p. 150-165, 2019.0925-2312http://hdl.handle.net/11449/18576810.1016/j.neucom.2019.04.070WOS:000470106400013Web of Sciencereponame:Repositório Institucional da UNESPinstname:Universidade Estadual Paulista (UNESP)instacron:UNESPengNeurocomputinginfo:eu-repo/semantics/openAccess2021-10-22T19:03:21Zoai:repositorio.unesp.br:11449/185768Repositório InstitucionalPUBhttp://repositorio.unesp.br/oai/requestopendoar:29462021-10-22T19:03:21Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP)false |
dc.title.none.fl_str_mv |
Combining clustering and active learning for the detection and learning of new image classes |
title |
Combining clustering and active learning for the detection and learning of new image classes |
spellingShingle |
Combining clustering and active learning for the detection and learning of new image classes Coletta, Luiz F. S. [UNESP] Image classification Active learning Clustering Open set Deep learning |
title_short |
Combining clustering and active learning for the detection and learning of new image classes |
title_full |
Combining clustering and active learning for the detection and learning of new image classes |
title_fullStr |
Combining clustering and active learning for the detection and learning of new image classes |
title_full_unstemmed |
Combining clustering and active learning for the detection and learning of new image classes |
title_sort |
Combining clustering and active learning for the detection and learning of new image classes |
author |
Coletta, Luiz F. S. [UNESP] |
author_facet |
Coletta, Luiz F. S. [UNESP] Ponti, Moacir Hruschka, Eduardo R. Acharya, Ayan Ghosh, Joydeep |
author_role |
author |
author2 |
Ponti, Moacir Hruschka, Eduardo R. Acharya, Ayan Ghosh, Joydeep |
author2_role |
author author author author |
dc.contributor.none.fl_str_mv |
Universidade Estadual Paulista (Unesp) Universidade de São Paulo (USP) Univ Texas Austin |
dc.contributor.author.fl_str_mv |
Coletta, Luiz F. S. [UNESP] Ponti, Moacir Hruschka, Eduardo R. Acharya, Ayan Ghosh, Joydeep |
dc.subject.por.fl_str_mv |
Image classification Active learning Clustering Open set Deep learning |
topic |
Image classification Active learning Clustering Open set Deep learning |
description |
Discriminative classification models often assume all classes are available at the training phase. As such models do not have a strategy to learn new concepts from available unlabeled instances, they usually work poorly when unknown classes emerge from future data to be classified. To address the appearance of new classes, some authors have developed approaches to transfer knowledge from known to unknown classes. Our study provides a more flexible approach to learn new (visual) classes that emerge over time. The key idea is materialized by an iterative classifier that combines Support Vector Machines with clustering via an optimization algorithm. An entropy and density-based selection strategy explores label uncertainty and high-density regions from unlabeled data to be classified. Selected instances from new classes are submitted to get labels and then used to improve the model. The proposed image classifier is consistently better than approaches that select instances randomly or from clusters. We also show that features obtained via Deep Learning methods improve results when compared with shallow features, but only using our selection strategy. Our approach requires fewer iterations to learn new classes, thereby significantly saving labeling costs. (C) 2019 Elsevier B.V. All rights reserved. |
publishDate |
2019 |
dc.date.none.fl_str_mv |
2019-10-04T12:38:21Z 2019-10-04T12:38:21Z 2019-09-17 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/article |
format |
article |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://dx.doi.org/10.1016/j.neucom.2019.04.070 Neurocomputing. Amsterdam: Elsevier, v. 358, p. 150-165, 2019. 0925-2312 http://hdl.handle.net/11449/185768 10.1016/j.neucom.2019.04.070 WOS:000470106400013 |
url |
http://dx.doi.org/10.1016/j.neucom.2019.04.070 http://hdl.handle.net/11449/185768 |
identifier_str_mv |
Neurocomputing. Amsterdam: Elsevier, v. 358, p. 150-165, 2019. 0925-2312 10.1016/j.neucom.2019.04.070 WOS:000470106400013 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.relation.none.fl_str_mv |
Neurocomputing |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
150-165 |
dc.publisher.none.fl_str_mv |
Elsevier B.V. |
publisher.none.fl_str_mv |
Elsevier B.V. |
dc.source.none.fl_str_mv |
Web of Science reponame:Repositório Institucional da UNESP instname:Universidade Estadual Paulista (UNESP) instacron:UNESP |
instname_str |
Universidade Estadual Paulista (UNESP) |
instacron_str |
UNESP |
institution |
UNESP |
reponame_str |
Repositório Institucional da UNESP |
collection |
Repositório Institucional da UNESP |
repository.name.fl_str_mv |
Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP) |
repository.mail.fl_str_mv |
|
_version_ |
1799965167256076288 |