A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images
Autor(a) principal: | |
---|---|
Data de Publicação: | 2020 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UFPE |
Texto Completo: | https://repositorio.ufpe.br/handle/123456789/39491 |
Resumo: | Hand pose estimation is an important task in computer vision due to its various fields of application, but mainly for providing a natural interaction between humans and machines. There are significant challenges for solving this task, primarily due to the high degree of freedom that is present in the human hand, and the possibility of self-occlusion. We investigate the usage of depthwise separable convolutions, an optimized convolution operation, to speed-up the inference time for convolutional models trained for 3D hand pose estimation. We show that the execution time for this approach can be improved to be up to 34.28% faster, while maintaining the accuracy scores on the metrics proposed by the literature. Additionally, we performed an extensive exploration and analysis of the use of depthwise separable convolutions regarding common challenges in tracking such as blur and noise, aiming to understand better in which scenarios this type of convolution impacts on the tracker precision. |
id |
UFPE_2f7e04af24144a1bb908943d0b132c1d |
---|---|
oai_identifier_str |
oai:repositorio.ufpe.br:123456789/39491 |
network_acronym_str |
UFPE |
network_name_str |
Repositório Institucional da UFPE |
repository_id_str |
2221 |
spelling |
COSTA, Willams de Limahttp://lattes.cnpq.br/3506588026663701http://lattes.cnpq.br/3355338790654065TEICHRIEB, Veronica2021-03-26T16:01:38Z2021-03-26T16:01:38Z2020-02-28ALVES, Thayonara de Pontes. A comprehensive exploration of depthwise separable convolutions for realtime 3D hand pose estimation through RGB images. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020.https://repositorio.ufpe.br/handle/123456789/39491Hand pose estimation is an important task in computer vision due to its various fields of application, but mainly for providing a natural interaction between humans and machines. There are significant challenges for solving this task, primarily due to the high degree of freedom that is present in the human hand, and the possibility of self-occlusion. We investigate the usage of depthwise separable convolutions, an optimized convolution operation, to speed-up the inference time for convolutional models trained for 3D hand pose estimation. We show that the execution time for this approach can be improved to be up to 34.28% faster, while maintaining the accuracy scores on the metrics proposed by the literature. Additionally, we performed an extensive exploration and analysis of the use of depthwise separable convolutions regarding common challenges in tracking such as blur and noise, aiming to understand better in which scenarios this type of convolution impacts on the tracker precision.FACEPEEstimação de pose de mãos é uma tarefa importante na visão computacional pelos seus vários campos de aplicação, mas principalmente por prover uma interação natural entre hu manos e máquinas. Existem desafios significativos para resolver essa tarefa, principalmente devido ao alto grau de liberdade que está presente na mão humana e a possibilidade de auto oclusão. Nós investigamos o uso de convoluções separáveis em profundidade, uma operação de convolução otimizada, para acelerar o tempo de inferência para modelos convolucionais treinados para estimação de pose 3D de mão. Nós mostramos que o tempo de execução para essa abordagem pode ser melhorado para ser até 34,28% mais rápido, mantendo os resulta dos de acurácia nas métricas propostas pela literatura. Adicionalmente, nós realizamos uma extensiva exploração e análise do uso de convoluções separáveis em profundidade em relação a desafios comuns em rastreamento, como borramento e ruído, procurando entender melhor em quais cenários esse tipo de convolução impacta na precisão do rastreador.engUniversidade Federal de PernambucoPrograma de Pos Graduacao em Ciencia da ComputacaoUFPEBrasilhttp://creativecommons.org/licenses/by-nc-nd/3.0/br/info:eu-repo/semantics/embargoedAccessMídia e interaçãoRedes neuraisA comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB imagesinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesismestradoreponame:Repositório Institucional da UFPEinstname:Universidade Federal de Pernambuco (UFPE)instacron:UFPETEXTDISSERTAÇÃO Willams de Lima Costa.pdf.txtDISSERTAÇÃO Willams de Lima Costa.pdf.txtExtracted texttext/plain157035https://repositorio.ufpe.br/bitstream/123456789/39491/4/DISSERTA%c3%87%c3%83O%20Willams%20de%20Lima%20Costa.pdf.txt1a7d4d02273a355dd8e3450b5195235aMD54THUMBNAILDISSERTAÇÃO Willams de Lima Costa.pdf.jpgDISSERTAÇÃO Willams de Lima Costa.pdf.jpgGenerated Thumbnailimage/jpeg1243https://repositorio.ufpe.br/bitstream/123456789/39491/5/DISSERTA%c3%87%c3%83O%20Willams%20de%20Lima%20Costa.pdf.jpgaca42d9e2447b8c261bcf9b1f530c6a9MD55LICENSElicense.txtlicense.txttext/plain; charset=utf-82310https://repositorio.ufpe.br/bitstream/123456789/39491/3/license.txtbd573a5ca8288eb7272482765f819534MD53ORIGINALDISSERTAÇÃO Willams de Lima Costa.pdfDISSERTAÇÃO Willams de Lima Costa.pdfapplication/pdf1829894https://repositorio.ufpe.br/bitstream/123456789/39491/1/DISSERTA%c3%87%c3%83O%20Willams%20de%20Lima%20Costa.pdf614f26ebb745d3995c4984aa1a2b8dc4MD51CC-LICENSElicense_rdflicense_rdfapplication/rdf+xml; charset=utf-8811https://repositorio.ufpe.br/bitstream/123456789/39491/2/license_rdfe39d27027a6cc9cb039ad269a5db8e34MD52123456789/394912021-03-27 02:15:39.256oai:repositorio.ufpe.br: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ório InstitucionalPUBhttps://repositorio.ufpe.br/oai/requestattena@ufpe.bropendoar:22212021-03-27T05:15:39Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE)false |
dc.title.pt_BR.fl_str_mv |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
title |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
spellingShingle |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images COSTA, Willams de Lima Mídia e interação Redes neurais |
title_short |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
title_full |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
title_fullStr |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
title_full_unstemmed |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
title_sort |
A comprehensive exploration of depthwise separable convolutions for real time 3D hand pose estimation through RGB images |
author |
COSTA, Willams de Lima |
author_facet |
COSTA, Willams de Lima |
author_role |
author |
dc.contributor.authorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/3506588026663701 |
dc.contributor.advisorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/3355338790654065 |
dc.contributor.author.fl_str_mv |
COSTA, Willams de Lima |
dc.contributor.advisor1.fl_str_mv |
TEICHRIEB, Veronica |
contributor_str_mv |
TEICHRIEB, Veronica |
dc.subject.por.fl_str_mv |
Mídia e interação Redes neurais |
topic |
Mídia e interação Redes neurais |
description |
Hand pose estimation is an important task in computer vision due to its various fields of application, but mainly for providing a natural interaction between humans and machines. There are significant challenges for solving this task, primarily due to the high degree of freedom that is present in the human hand, and the possibility of self-occlusion. We investigate the usage of depthwise separable convolutions, an optimized convolution operation, to speed-up the inference time for convolutional models trained for 3D hand pose estimation. We show that the execution time for this approach can be improved to be up to 34.28% faster, while maintaining the accuracy scores on the metrics proposed by the literature. Additionally, we performed an extensive exploration and analysis of the use of depthwise separable convolutions regarding common challenges in tracking such as blur and noise, aiming to understand better in which scenarios this type of convolution impacts on the tracker precision. |
publishDate |
2020 |
dc.date.issued.fl_str_mv |
2020-02-28 |
dc.date.accessioned.fl_str_mv |
2021-03-26T16:01:38Z |
dc.date.available.fl_str_mv |
2021-03-26T16:01:38Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.citation.fl_str_mv |
ALVES, Thayonara de Pontes. A comprehensive exploration of depthwise separable convolutions for realtime 3D hand pose estimation through RGB images. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020. |
dc.identifier.uri.fl_str_mv |
https://repositorio.ufpe.br/handle/123456789/39491 |
identifier_str_mv |
ALVES, Thayonara de Pontes. A comprehensive exploration of depthwise separable convolutions for realtime 3D hand pose estimation through RGB images. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020. |
url |
https://repositorio.ufpe.br/handle/123456789/39491 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
http://creativecommons.org/licenses/by-nc-nd/3.0/br/ info:eu-repo/semantics/embargoedAccess |
rights_invalid_str_mv |
http://creativecommons.org/licenses/by-nc-nd/3.0/br/ |
eu_rights_str_mv |
embargoedAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.publisher.program.fl_str_mv |
Programa de Pos Graduacao em Ciencia da Computacao |
dc.publisher.initials.fl_str_mv |
UFPE |
dc.publisher.country.fl_str_mv |
Brasil |
publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFPE instname:Universidade Federal de Pernambuco (UFPE) instacron:UFPE |
instname_str |
Universidade Federal de Pernambuco (UFPE) |
instacron_str |
UFPE |
institution |
UFPE |
reponame_str |
Repositório Institucional da UFPE |
collection |
Repositório Institucional da UFPE |
bitstream.url.fl_str_mv |
https://repositorio.ufpe.br/bitstream/123456789/39491/4/DISSERTA%c3%87%c3%83O%20Willams%20de%20Lima%20Costa.pdf.txt https://repositorio.ufpe.br/bitstream/123456789/39491/5/DISSERTA%c3%87%c3%83O%20Willams%20de%20Lima%20Costa.pdf.jpg https://repositorio.ufpe.br/bitstream/123456789/39491/3/license.txt https://repositorio.ufpe.br/bitstream/123456789/39491/1/DISSERTA%c3%87%c3%83O%20Willams%20de%20Lima%20Costa.pdf https://repositorio.ufpe.br/bitstream/123456789/39491/2/license_rdf |
bitstream.checksum.fl_str_mv |
1a7d4d02273a355dd8e3450b5195235a aca42d9e2447b8c261bcf9b1f530c6a9 bd573a5ca8288eb7272482765f819534 614f26ebb745d3995c4984aa1a2b8dc4 e39d27027a6cc9cb039ad269a5db8e34 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 MD5 MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE) |
repository.mail.fl_str_mv |
attena@ufpe.br |
_version_ |
1802310638717894656 |