MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING

Detalhes bibliográficos
Autor(a) principal: Moulin, Jordão Cabral
Data de Publicação: 2022
Outros Autores: Lopes, Dercilio Junior Verly, Mulin, Lucas Braga, Bobadilha, Gabrielly dos Santos, Oliveira, Ramon Ferreir
Tipo de documento: Artigo
Idioma: eng
Título da fonte: Cerne (Online)
Texto Completo: https://cerne.ufla.br/site/index.php/CERNE/article/view/2978
Resumo: Background: Multiple challenges are faced by industry and certification agencies when commercializing tropical species. Anatomical similarities of tropical hardwoods impair identification. Deep learning models can facilitate microscopic identification of wood by using sophisticated techniques such as deep convolutional networks (DCNN). Our objective was to microscopically identify 23 wood species using a custom DCNN model. Results: Photographs from microscopic slides of each wood species were processed, and the final data set contained 2,448 images. We applied stratified k-fold cross-validation technique during training to increase model’s robustness and trustworthiness. Thus, the dataset was divided into approximately 80% training (1,958 images) and 20% validation (490 images) for each fold. A series of augmentations were performed only on training data to include variations in rotation, zoom, and perspective. Image augmentation was performed on-the-fly. The network consisted of convolutions, max pooling, global average pooling, and fully connected layers. We tested the performance of the DCNN against accuracy, precision, recall, and F1-score on the validation set for each fold. Conclusion: The machine learned custom model accuracy was considered excellent (>0.90). The model’s worst performance was identified in distinguishing between Toona ciliata and Khaya ivorensis, which was due more to wood variability than to a machine learning deficiency. Future studies should focus on integration, verification/monitoring, and updating of current models for end user manipulation, trust, ethics, and security.
id UFLA-3_97d122105dd7978ff1b8e887e5060d0f
oai_identifier_str oai:cerne.ufla.br:article/2978
network_acronym_str UFLA-3
network_name_str Cerne (Online)
repository_id_str
spelling MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNINGconvolutional neural networks (CNN)deep learningtropical specieswood anatomywood identificationBackground: Multiple challenges are faced by industry and certification agencies when commercializing tropical species. Anatomical similarities of tropical hardwoods impair identification. Deep learning models can facilitate microscopic identification of wood by using sophisticated techniques such as deep convolutional networks (DCNN). Our objective was to microscopically identify 23 wood species using a custom DCNN model. Results: Photographs from microscopic slides of each wood species were processed, and the final data set contained 2,448 images. We applied stratified k-fold cross-validation technique during training to increase model’s robustness and trustworthiness. Thus, the dataset was divided into approximately 80% training (1,958 images) and 20% validation (490 images) for each fold. A series of augmentations were performed only on training data to include variations in rotation, zoom, and perspective. Image augmentation was performed on-the-fly. The network consisted of convolutions, max pooling, global average pooling, and fully connected layers. We tested the performance of the DCNN against accuracy, precision, recall, and F1-score on the validation set for each fold. Conclusion: The machine learned custom model accuracy was considered excellent (>0.90). The model’s worst performance was identified in distinguishing between Toona ciliata and Khaya ivorensis, which was due more to wood variability than to a machine learning deficiency. Future studies should focus on integration, verification/monitoring, and updating of current models for end user manipulation, trust, ethics, and security.CERNECERNE2022-08-05info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionapplication/pdfhttps://cerne.ufla.br/site/index.php/CERNE/article/view/2978CERNE; Vol 28 No 1 (2022); e-102978CERNE; Vol 28 No 1 (2022); e-1029782317-63420104-7760reponame:Cerne (Online)instname:Universidade Federal de Lavras (UFLA)instacron:UFLAenghttps://cerne.ufla.br/site/index.php/CERNE/article/view/2978/1295http://creativecommons.org/licenses/by/4.0/info:eu-repo/semantics/openAccessMoulin, Jordão CabralLopes, Dercilio Junior VerlyMulin, Lucas BragaBobadilha, Gabrielly dos SantosOliveira, Ramon Ferreir2022-08-25T17:35:35Zoai:cerne.ufla.br:article/2978Revistahttps://cerne.ufla.br/site/index.php/CERNEPUBhttps://cerne.ufla.br/site/index.php/CERNE/oaicerne@dcf.ufla.br||cerne@dcf.ufla.br2317-63420104-7760opendoar:2024-05-21T19:54:48.225515Cerne (Online) - Universidade Federal de Lavras (UFLA)true
dc.title.none.fl_str_mv MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
title MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
spellingShingle MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
Moulin, Jordão Cabral
convolutional neural networks (CNN)
deep learning
tropical species
wood anatomy
wood identification
title_short MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
title_full MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
title_fullStr MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
title_full_unstemmed MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
title_sort MICROSCOPIC IDENTIFICATION OF BRAZILIAN COMMERCIAL WOOD SPECIES VIA MACHINE-LEARNING
author Moulin, Jordão Cabral
author_facet Moulin, Jordão Cabral
Lopes, Dercilio Junior Verly
Mulin, Lucas Braga
Bobadilha, Gabrielly dos Santos
Oliveira, Ramon Ferreir
author_role author
author2 Lopes, Dercilio Junior Verly
Mulin, Lucas Braga
Bobadilha, Gabrielly dos Santos
Oliveira, Ramon Ferreir
author2_role author
author
author
author
dc.contributor.author.fl_str_mv Moulin, Jordão Cabral
Lopes, Dercilio Junior Verly
Mulin, Lucas Braga
Bobadilha, Gabrielly dos Santos
Oliveira, Ramon Ferreir
dc.subject.por.fl_str_mv convolutional neural networks (CNN)
deep learning
tropical species
wood anatomy
wood identification
topic convolutional neural networks (CNN)
deep learning
tropical species
wood anatomy
wood identification
description Background: Multiple challenges are faced by industry and certification agencies when commercializing tropical species. Anatomical similarities of tropical hardwoods impair identification. Deep learning models can facilitate microscopic identification of wood by using sophisticated techniques such as deep convolutional networks (DCNN). Our objective was to microscopically identify 23 wood species using a custom DCNN model. Results: Photographs from microscopic slides of each wood species were processed, and the final data set contained 2,448 images. We applied stratified k-fold cross-validation technique during training to increase model’s robustness and trustworthiness. Thus, the dataset was divided into approximately 80% training (1,958 images) and 20% validation (490 images) for each fold. A series of augmentations were performed only on training data to include variations in rotation, zoom, and perspective. Image augmentation was performed on-the-fly. The network consisted of convolutions, max pooling, global average pooling, and fully connected layers. We tested the performance of the DCNN against accuracy, precision, recall, and F1-score on the validation set for each fold. Conclusion: The machine learned custom model accuracy was considered excellent (>0.90). The model’s worst performance was identified in distinguishing between Toona ciliata and Khaya ivorensis, which was due more to wood variability than to a machine learning deficiency. Future studies should focus on integration, verification/monitoring, and updating of current models for end user manipulation, trust, ethics, and security.
publishDate 2022
dc.date.none.fl_str_mv 2022-08-05
dc.type.driver.fl_str_mv info:eu-repo/semantics/article
info:eu-repo/semantics/publishedVersion
format article
status_str publishedVersion
dc.identifier.uri.fl_str_mv https://cerne.ufla.br/site/index.php/CERNE/article/view/2978
url https://cerne.ufla.br/site/index.php/CERNE/article/view/2978
dc.language.iso.fl_str_mv eng
language eng
dc.relation.none.fl_str_mv https://cerne.ufla.br/site/index.php/CERNE/article/view/2978/1295
dc.rights.driver.fl_str_mv http://creativecommons.org/licenses/by/4.0/
info:eu-repo/semantics/openAccess
rights_invalid_str_mv http://creativecommons.org/licenses/by/4.0/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv CERNE
CERNE
publisher.none.fl_str_mv CERNE
CERNE
dc.source.none.fl_str_mv CERNE; Vol 28 No 1 (2022); e-102978
CERNE; Vol 28 No 1 (2022); e-102978
2317-6342
0104-7760
reponame:Cerne (Online)
instname:Universidade Federal de Lavras (UFLA)
instacron:UFLA
instname_str Universidade Federal de Lavras (UFLA)
instacron_str UFLA
institution UFLA
reponame_str Cerne (Online)
collection Cerne (Online)
repository.name.fl_str_mv Cerne (Online) - Universidade Federal de Lavras (UFLA)
repository.mail.fl_str_mv cerne@dcf.ufla.br||cerne@dcf.ufla.br
_version_ 1799874944275841024