An efficient action detection from first person vision with attention model

Autores
Straminsky, Axel; Jacobo, Julio; Buemi, María Elena
Año de publicación
2021
Idioma
inglés
Tipo de recurso
documento de conferencia
Estado
versión publicada
Descripción
The goal of this work is to propose possible improvements on one of the latest models for Video Action Recognition based on currently existing attention mechanisms. We took a model architecture that uses 2 sub-models in paralell: one based on Optical Flow and the other based on the video itself, and proposed the following improvements: adding mixed precision in the training loop, using a Ranger optimizer instead of SGD, and expanding the Attention Mechanism. The video database used for this work was the EGTEA+ that is a action database of first person videos of daily activities.
Sociedad Argentina de Informática e Investigación Operativa
Materia
Ciencias Informáticas
First Person Vision
Human Computer Interaction
Action Recognition
Attention module
Nivel de accesibilidad
acceso abierto
Condiciones de uso
http://creativecommons.org/licenses/by-nc-sa/3.0/
Repositorio
SEDICI (UNLP)
Institución
Universidad Nacional de La Plata
OAI Identificador
oai:sedici.unlp.edu.ar:10915/141291

id SEDICI_a95be7c5464ddeaa4f1a48013637a885
oai_identifier_str oai:sedici.unlp.edu.ar:10915/141291
network_acronym_str SEDICI
repository_id_str 1329
network_name_str SEDICI (UNLP)
spelling An efficient action detection from first person vision with attention modelStraminsky, AxelJacobo, JulioBuemi, María ElenaCiencias InformáticasFirst Person VisionHuman Computer InteractionAction RecognitionAttention moduleThe goal of this work is to propose possible improvements on one of the latest models for Video Action Recognition based on currently existing attention mechanisms. We took a model architecture that uses 2 sub-models in paralell: one based on Optical Flow and the other based on the video itself, and proposed the following improvements: adding mixed precision in the training loop, using a Ranger optimizer instead of SGD, and expanding the Attention Mechanism. The video database used for this work was the EGTEA+ that is a action database of first person videos of daily activities.Sociedad Argentina de Informática e Investigación Operativa2021-10info:eu-repo/semantics/conferenceObjectinfo:eu-repo/semantics/publishedVersionObjeto de conferenciahttp://purl.org/coar/resource_type/c_5794info:ar-repo/semantics/documentoDeConferenciaapplication/pdf36-39http://sedici.unlp.edu.ar/handle/10915/141291enginfo:eu-repo/semantics/altIdentifier/url/http://50jaiio.sadio.org.ar/pdfs/saiv/SAIV-08.pdfinfo:eu-repo/semantics/altIdentifier/issn/2683-8990info:eu-repo/semantics/openAccesshttp://creativecommons.org/licenses/by-nc-sa/3.0/Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)reponame:SEDICI (UNLP)instname:Universidad Nacional de La Platainstacron:UNLP2025-09-29T11:35:53Zoai:sedici.unlp.edu.ar:10915/141291Institucionalhttp://sedici.unlp.edu.ar/Universidad públicaNo correspondehttp://sedici.unlp.edu.ar/oai/snrdalira@sedici.unlp.edu.arArgentinaNo correspondeNo correspondeNo correspondeopendoar:13292025-09-29 11:35:53.84SEDICI (UNLP) - Universidad Nacional de La Platafalse
dc.title.none.fl_str_mv An efficient action detection from first person vision with attention model
title An efficient action detection from first person vision with attention model
spellingShingle An efficient action detection from first person vision with attention model
Straminsky, Axel
Ciencias Informáticas
First Person Vision
Human Computer Interaction
Action Recognition
Attention module
title_short An efficient action detection from first person vision with attention model
title_full An efficient action detection from first person vision with attention model
title_fullStr An efficient action detection from first person vision with attention model
title_full_unstemmed An efficient action detection from first person vision with attention model
title_sort An efficient action detection from first person vision with attention model
dc.creator.none.fl_str_mv Straminsky, Axel
Jacobo, Julio
Buemi, María Elena
author Straminsky, Axel
author_facet Straminsky, Axel
Jacobo, Julio
Buemi, María Elena
author_role author
author2 Jacobo, Julio
Buemi, María Elena
author2_role author
author
dc.subject.none.fl_str_mv Ciencias Informáticas
First Person Vision
Human Computer Interaction
Action Recognition
Attention module
topic Ciencias Informáticas
First Person Vision
Human Computer Interaction
Action Recognition
Attention module
dc.description.none.fl_txt_mv The goal of this work is to propose possible improvements on one of the latest models for Video Action Recognition based on currently existing attention mechanisms. We took a model architecture that uses 2 sub-models in paralell: one based on Optical Flow and the other based on the video itself, and proposed the following improvements: adding mixed precision in the training loop, using a Ranger optimizer instead of SGD, and expanding the Attention Mechanism. The video database used for this work was the EGTEA+ that is a action database of first person videos of daily activities.
Sociedad Argentina de Informática e Investigación Operativa
description The goal of this work is to propose possible improvements on one of the latest models for Video Action Recognition based on currently existing attention mechanisms. We took a model architecture that uses 2 sub-models in paralell: one based on Optical Flow and the other based on the video itself, and proposed the following improvements: adding mixed precision in the training loop, using a Ranger optimizer instead of SGD, and expanding the Attention Mechanism. The video database used for this work was the EGTEA+ that is a action database of first person videos of daily activities.
publishDate 2021
dc.date.none.fl_str_mv 2021-10
dc.type.none.fl_str_mv info:eu-repo/semantics/conferenceObject
info:eu-repo/semantics/publishedVersion
Objeto de conferencia
http://purl.org/coar/resource_type/c_5794
info:ar-repo/semantics/documentoDeConferencia
format conferenceObject
status_str publishedVersion
dc.identifier.none.fl_str_mv http://sedici.unlp.edu.ar/handle/10915/141291
url http://sedici.unlp.edu.ar/handle/10915/141291
dc.language.none.fl_str_mv eng
language eng
dc.relation.none.fl_str_mv info:eu-repo/semantics/altIdentifier/url/http://50jaiio.sadio.org.ar/pdfs/saiv/SAIV-08.pdf
info:eu-repo/semantics/altIdentifier/issn/2683-8990
dc.rights.none.fl_str_mv info:eu-repo/semantics/openAccess
http://creativecommons.org/licenses/by-nc-sa/3.0/
Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)
eu_rights_str_mv openAccess
rights_invalid_str_mv http://creativecommons.org/licenses/by-nc-sa/3.0/
Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)
dc.format.none.fl_str_mv application/pdf
36-39
dc.source.none.fl_str_mv reponame:SEDICI (UNLP)
instname:Universidad Nacional de La Plata
instacron:UNLP
reponame_str SEDICI (UNLP)
collection SEDICI (UNLP)
instname_str Universidad Nacional de La Plata
instacron_str UNLP
institution UNLP
repository.name.fl_str_mv SEDICI (UNLP) - Universidad Nacional de La Plata
repository.mail.fl_str_mv alira@sedici.unlp.edu.ar
_version_ 1844616237421690880
score 13.070432