Logo do repositório
 
Publicação

Continual learning for object classification: integrating AutoML for binary classification tasks within a modular dynamic architecture

2024Artigo científico Acesso aberto
Ver registo simples
datacite.subject.sdg09:Indústria, Inovação e Infraestruturas
datacite.subject.sdg04:Educação de Qualidade
datacite.subject.sdg08:Trabalho Digno e Crescimento Económico
dc.contributor.authorTurner, Daniel
dc.contributor.authorCardoso, Pedro
dc.contributor.authorRodrigues, Joao
dc.date.accessioned2026-04-29T12:42:06Z
dc.date.available2026-04-29T12:42:06Z
dc.date.issued2024
dc.description.abstractFor humans it is quite easy to identify a new object after learning to identify existing ones, but not for a machine. Deep neural networks (DNN) are the foundation of the current state-of-the-art methods for training machines to recognize sets of objects. The issue is that any modification to the DNN weights that were trained to classify an initial set of objects has the potential to seriously impair the network’s ability to make those initial classifications; this behaviour is referred to as catastrophic forgetting (CF). This paper presents a continual learning (CL) architecture that can deal with CF. The architecture is composed of two primary parts: (i) The feature extraction component, which is based on the ResNet50 backbone and (ii) the modular dynamic classification (MDC) component. The latter is made up of multiple sub-networks that gradually assemble themselves into a tree-like structure that reorganizes itself as it learns over time, so that each sub-network can operate independently. The MDC relies heavily on binary classification, and here the application of automated machine learning (AutoML) was introduced, where each binary classifier is tailored on-the-fly, and is/can be different from object to object. The strategy involves a calculated selection from a predefined list of model types and parameters, optimizing them for their respective tasks. Results demonstrate that we advanced the adaptability and performance of the network, emphasizing the transformative potential of AutoML in modular CL approaches. Tests on the CORe50 dataset showed accuracy results of 81.1%, which are above the state of the art for CL architectures.eng
dc.description.sponsorshipUIDP/04516/2020
dc.identifier.doi10.1109/access.2024.3510536
dc.identifier.issn2169-3536
dc.identifier.urihttp://hdl.handle.net/10400.1/28804
dc.language.isoeng
dc.peerreviewedyes
dc.publisherInstitute of Electrical and Electronics Engineers
dc.relationNOVA Laboratory for Computer Science and Informatics
dc.relation.ispartofIEEE Access
dc.rights.urihttp://creativecommons.org/licenses/by/4.0/
dc.subjectAutoML
dc.subjectBinary classification
dc.subjectCatastrophic forgetting
dc.subjectComputer vision
dc.subjectContinual learning
dc.subjectNeural networks
dc.subjectObject recognition
dc.titleContinual learning for object classification: integrating AutoML for binary classification tasks within a modular dynamic architectureeng
dc.typejournal article
dspace.entity.typePublication
oaire.awardNumberUIDB/04516/2020
oaire.awardTitleNOVA Laboratory for Computer Science and Informatics
oaire.awardURIinfo:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDB%2F04516%2F2020/PT
oaire.citation.endPage183742
oaire.citation.startPage183725
oaire.citation.titleIEEE Access
oaire.citation.volume12
oaire.fundingStream6817 - DCRRNI ID
oaire.versionhttp://purl.org/coar/version/c_970fb48d4fbd8a85
person.familyNameTurner
person.familyNameCardoso
person.familyNameRodrigues
person.givenNameDaniel
person.givenNamePedro
person.givenNameJoao
person.identifier.ciencia-id5F10-1C37-FE45
person.identifier.ciencia-id8A19-98F7-9914
person.identifier.orcid0000-0002-1198-9841
person.identifier.orcid0000-0003-4803-7964
person.identifier.orcid0000-0002-3562-6025
person.identifier.ridG-6405-2013
person.identifier.scopus-author-id35602693500
person.identifier.scopus-author-id55807461600
project.funder.identifierhttp://doi.org/10.13039/501100001871
project.funder.nameFundação para a Ciência e a Tecnologia
relation.isAuthorOfPublicationad5a3f8c-533d-4dd8-95a5-0385df6441f2
relation.isAuthorOfPublication62bebc54-51ee-4e35-bcf5-6dd69efd09e0
relation.isAuthorOfPublication683ba85b-459c-4789-a4ff-a4e2a904b295
relation.isAuthorOfPublication.latestForDiscoveryad5a3f8c-533d-4dd8-95a5-0385df6441f2
relation.isProjectOfPublication1122b3d4-9740-4ad7-9abf-86bb7a3615da
relation.isProjectOfPublication.latestForDiscovery1122b3d4-9740-4ad7-9abf-86bb7a3615da

Ficheiros

Principais
A mostrar 1 - 1 de 1
Miniatura
Nome:
Continual_Learning_for_Object_Classification_Integrating_AutoML_for_Binary_Classification_Tasks_Within_a_Modular_Dynamic_Architecture.pdf
Tamanho:
3.17 MB
Formato:
Adobe Portable Document Format
Ver/Abrir
Licença
A mostrar 1 - 1 de 1
Miniatura indisponível
Nome:
license.txt
Tamanho:
3.46 KB
Formato:
Item-specific license agreed upon to submission
Descrição:
Ver/Abrir

Coleções

FCT2-Artigos (em revistas ou actas indexadas)