Deep ensemble-based hard sample mining for food recognition

Deep neural networks represent a compelling technique to tackle complex real-world problems, but are over-parameterized and often suffer from over- or under-confident estimates. Deep ensembles have shown better parameter estimations and often provide reliable uncertainty estimates that contribute to...

Full description

Bibliographic Details
Authors: Nagarajan, Bhalaji, Bolaños Solà, Marc, Aguilar Torres, Eduardo, Radeva, Petia
Format: article
Status:Published version
Publication Date:2023
Country:España
Institution:Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya)
Repository:Recercat. Dipósit de la Recerca de Catalunya
OAI Identifier:oai:recercat.cat:2445/219021
Online Access:https://hdl.handle.net/2445/219021
Access Level:Open access
Keyword:Aprenentatge automàtic
Reconeixement de formes (Informàtica)
Visió per ordinador
Machine learning
Pattern recognition systems
Computer vision
id ES_c9edfb3d68beeda5d45b1f5f5262b3fe
oai_identifier_str oai:recercat.cat:2445/219021
network_acronym_str ES
network_name_str España
repository_id_str
spelling Deep ensemble-based hard sample mining for food recognitionNagarajan, BhalajiBolaños Solà, MarcAguilar Torres, EduardoRadeva, PetiaAprenentatge automàticReconeixement de formes (Informàtica)Visió per ordinadorMachine learningPattern recognition systemsComputer visionDeep neural networks represent a compelling technique to tackle complex real-world problems, but are over-parameterized and often suffer from over- or under-confident estimates. Deep ensembles have shown better parameter estimations and often provide reliable uncertainty estimates that contribute to the robustness of the results. In this work, we propose a new metric to identify samples that are hard to classify. Our metric is defined as coincidence score for deep ensembles which measures the agreement of its individual models. The main hypothesis we rely on is that deep learning algorithms learn the low-loss samples better compared to large-loss samples. In order to compensate for this, we use controlled over-sampling on the identified ”hard” samples using proper data augmentation schemes to enable the models to learn those samples better. We validate the proposed metric using two public food datasets on different backbone architectures and show the improvements compared to the conventional deep neural network training using different performance metrics.Elsevier2025202520232025info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersion11 p.application/pdfhttps://hdl.handle.net/2445/219021Articles publicats en revistes (Matemàtiques i Informàtica)reponame:Recercat. Dipósit de la Recerca de Catalunyainstname:Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya)InglésReproducció del document publicat a: https://doi.org/10.1016/j.jvcir.2023.103905Journal of Visual Communication and Image Representation, 2023, vol. 95https://doi.org/10.1016/j.jvcir.2023.103905cc-by (c) Bhalaji Nagarajan, 2023http://creativecommons.org/licenses/by/3.0/es/info:eu-repo/semantics/openAccessoai:recercat.cat:2445/2190212026-05-29T05:05:01Z
dc.title.none.fl_str_mv Deep ensemble-based hard sample mining for food recognition
title Deep ensemble-based hard sample mining for food recognition
spellingShingle Deep ensemble-based hard sample mining for food recognition
Nagarajan, Bhalaji
Aprenentatge automàtic
Reconeixement de formes (Informàtica)
Visió per ordinador
Machine learning
Pattern recognition systems
Computer vision
title_short Deep ensemble-based hard sample mining for food recognition
title_full Deep ensemble-based hard sample mining for food recognition
title_fullStr Deep ensemble-based hard sample mining for food recognition
title_full_unstemmed Deep ensemble-based hard sample mining for food recognition
title_sort Deep ensemble-based hard sample mining for food recognition
dc.creator.none.fl_str_mv Nagarajan, Bhalaji
Bolaños Solà, Marc
Aguilar Torres, Eduardo
Radeva, Petia
author Nagarajan, Bhalaji
author_facet Nagarajan, Bhalaji
Bolaños Solà, Marc
Aguilar Torres, Eduardo
Radeva, Petia
author_role author
author2 Bolaños Solà, Marc
Aguilar Torres, Eduardo
Radeva, Petia
author2_role author
author
author
dc.subject.none.fl_str_mv Aprenentatge automàtic
Reconeixement de formes (Informàtica)
Visió per ordinador
Machine learning
Pattern recognition systems
Computer vision
topic Aprenentatge automàtic
Reconeixement de formes (Informàtica)
Visió per ordinador
Machine learning
Pattern recognition systems
Computer vision
description Deep neural networks represent a compelling technique to tackle complex real-world problems, but are over-parameterized and often suffer from over- or under-confident estimates. Deep ensembles have shown better parameter estimations and often provide reliable uncertainty estimates that contribute to the robustness of the results. In this work, we propose a new metric to identify samples that are hard to classify. Our metric is defined as coincidence score for deep ensembles which measures the agreement of its individual models. The main hypothesis we rely on is that deep learning algorithms learn the low-loss samples better compared to large-loss samples. In order to compensate for this, we use controlled over-sampling on the identified ”hard” samples using proper data augmentation schemes to enable the models to learn those samples better. We validate the proposed metric using two public food datasets on different backbone architectures and show the improvements compared to the conventional deep neural network training using different performance metrics.
publishDate 2023
dc.date.none.fl_str_mv 2023
2025
2025
2025
dc.type.none.fl_str_mv info:eu-repo/semantics/article
info:eu-repo/semantics/publishedVersion
format article
status_str publishedVersion
dc.identifier.none.fl_str_mv https://hdl.handle.net/2445/219021
url https://hdl.handle.net/2445/219021
dc.language.none.fl_str_mv Inglés
language_invalid_str_mv Inglés
dc.relation.none.fl_str_mv Reproducció del document publicat a: https://doi.org/10.1016/j.jvcir.2023.103905
Journal of Visual Communication and Image Representation, 2023, vol. 95
https://doi.org/10.1016/j.jvcir.2023.103905
dc.rights.none.fl_str_mv cc-by (c) Bhalaji Nagarajan, 2023
http://creativecommons.org/licenses/by/3.0/es/
info:eu-repo/semantics/openAccess
rights_invalid_str_mv cc-by (c) Bhalaji Nagarajan, 2023
http://creativecommons.org/licenses/by/3.0/es/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv 11 p.
application/pdf
dc.publisher.none.fl_str_mv Elsevier
publisher.none.fl_str_mv Elsevier
dc.source.none.fl_str_mv Articles publicats en revistes (Matemàtiques i Informàtica)
reponame:Recercat. Dipósit de la Recerca de Catalunya
instname:Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya)
instname_str Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya)
reponame_str Recercat. Dipósit de la Recerca de Catalunya
collection Recercat. Dipósit de la Recerca de Catalunya
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869419434419945472
score 15,812429

Similar Items