Utilize este identificador para referenciar este registo:
https://hdl.handle.net/1822/86333
Registo completo
Campo DC | Valor | Idioma |
---|---|---|
dc.contributor.author | Oliveira Santos, Flavio Arthur | por |
dc.contributor.author | Zanchettin, Cleber | por |
dc.contributor.author | Matos, Leonardo Nogueira | por |
dc.contributor.author | Novais, Paulo | por |
dc.date.accessioned | 2023-09-12T14:33:42Z | - |
dc.date.issued | 2022 | - |
dc.identifier.citation | Flávio Arthur Oliveira Santos, Cleber Zanchettin, Leonardo Nogueira Matos, Paulo Novais, On the Impact of Interpretability Methods in Active Image Augmentation Method, Logic Journal of the IGPL, Volume 30, Issue 4, August 2022, Pages 611–621, https://doi.org/10.1093/jigpal/jzab006 | por |
dc.identifier.issn | 1367-0751 | - |
dc.identifier.uri | https://hdl.handle.net/1822/86333 | - |
dc.description.abstract | Robustness is a significant constraint in machine learning models. The performance of the algorithms must not deteriorate when training and testing with slightly different data. Deep neural network models achieve awe-inspiring results in a wide range of applications of computer vision. Still, in the presence of noise or region occlusion, some models exhibit inaccurate performance even with data handled in training. Besides, some experiments suggest deep learning models sometimes use incorrect parts of the input information to perform inference. Active image augmentation (ADA) is an augmentation method that uses interpretability methods to augment the training data and improve its robustness to face the described problems. Although ADA presented interesting results, its original version only used the vanilla backpropagation interpretability to train the U-Net model. In this work, we propose an extensive experimental analysis of the interpretability method's impact on ADA. We use five interpretability methods: vanilla backpropagation, guided backpropagation, gradient-weighted class activation mapping (GradCam), guided GradCam and InputXGradient. The results show that all methods achieve similar performance at the ending of training, but when combining ADA with GradCam, the U-Net model presented an impressive fast convergence. | por |
dc.description.sponsorship | This work has been supported by FundacAo para a Ciencia e Tecnologia within the Project Scope: UIDB/00319/2020. The authors also thank CoordenacAo de Aperfeicoamento de Pessoal de Nivel Superior and Conselho Nacional de Desenvolvimento Cientifico e Tecnologico (Brazilian Researcher Agencies) for the financial support. | por |
dc.language.iso | eng | por |
dc.publisher | Oxford University Press | por |
dc.relation | info:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDB%2F00319%2F2020/PT | por |
dc.rights | restrictedAccess | por |
dc.subject | Data augmentation | por |
dc.subject | robustness | por |
dc.subject | interpretability | por |
dc.title | On the impact of interpretability methods in active image augmentation method | por |
dc.type | article | por |
dc.peerreviewed | yes | por |
dc.relation.publisherversion | https://academic.oup.com/jigpal/article/30/4/611/6123345 | por |
oaire.citationStartPage | 611 | por |
oaire.citationEndPage | 621 | por |
oaire.citationIssue | 4 | por |
oaire.citationVolume | 30 | por |
dc.date.updated | 2023-07-31T23:14:45Z | - |
dc.identifier.doi | 10.1093/jigpal/jzab006 | por |
dc.date.embargo | 10000-01-01 | - |
dc.subject.wos | Science & Technology | - |
sdum.export.identifier | 12656 | - |
sdum.journal | Logic Journal of the IGPL | por |
oaire.version | AM | por |
Aparece nas coleções: | CAlg - Artigos em revistas internacionais / Papers in international journals |
Ficheiros deste registo:
Ficheiro | Descrição | Tamanho | Formato | |
---|---|---|---|---|
2102.12354.pdf Acesso restrito! | 2,55 MB | Adobe PDF | Ver/Abrir |