Facial action unit detection with convolutional neural networks

We propose a novel deep convolutional neural network architecture to study the problem of action unit detection. We leverage recent gains in large-scale object recognition by formulating the task of predicting the presence of a specific action unit in a still image as simple image-level binary class...

Full description

Autores:
Romero Vergara, Andrés Felipe
Tipo de recurso:
Fecha de publicación:
2017
Institución:
Universidad de los Andes
Repositorio:
Séneca: repositorio Uniandes
Idioma:
eng
OAI Identifier:
oai:repositorio.uniandes.edu.co:1992/13826
Acceso en línea:
http://hdl.handle.net/1992/13826
Palabra clave:
Redes neurales (Computadores) - Investigaciones
Procesamiento de imágenes - Investigaciones
Sistemas de reconocimiento de configuraciones - Investigaciones
Expresión facial - Procesamiento de imágenes - Investigaciones
Ingeniería
Rights
openAccess
License
http://creativecommons.org/licenses/by-nc-nd/4.0/
id UNIANDES2_cff3d629fceee637bf405fbdfbf58ef0
oai_identifier_str oai:repositorio.uniandes.edu.co:1992/13826
network_acronym_str UNIANDES2
network_name_str Séneca: repositorio Uniandes
repository_id_str
spelling Al consultar y hacer uso de este recurso, está aceptando las condiciones de uso establecidas por los autores.http://creativecommons.org/licenses/by-nc-nd/4.0/info:eu-repo/semantics/openAccesshttp://purl.org/coar/access_right/c_abf2Arbeláez Escalante, Pablo Andrés7b73426f-f63b-413f-b44b-ddfa70416b65400Romero Vergara, Andrés Felipe103125002018-09-28T10:56:43Z2018-09-28T10:56:43Z2017http://hdl.handle.net/1992/13826u729464.pdfinstname:Universidad de los Andesreponame:Repositorio Institucional Sénecarepourl:https://repositorio.uniandes.edu.co/We propose a novel deep convolutional neural network architecture to study the problem of action unit detection. We leverage recent gains in large-scale object recognition by formulating the task of predicting the presence of a specific action unit in a still image as simple image-level binary classification. We first train a convolutional encoder on the problem of multi-view emotion recognition as a high-level representation of facial expressions. We show that our architecture generalizes across views, ethnicity, gender and age by merging and training jointly on three standard emotion recognition datasets: CK+, Bosphorus and RafD. Our system is the first fully multi-view emotion recognizer proposed in the literature. We then extend this shared learned representation with fully-connected layers trained to detect individual action units. Our approach is conceptually simpler and yet significantly more accurate than the best methods based on the dominant paradigm for the study of this problem, which relies on facial landmark detection as an intermediate task. We conduct experiments on the BP4D dataset, the largest and most challenging benchmark currently available for action unit detection, and report an absolute improvement of 16% over the previous state-of-the-art.Magíster en Ingeniería BiomédicaMaestría19 hojasapplication/pdfengUniandesMaestría en Ingeniería BiomédicaFacultad de IngenieríaDepartamento de Ingeniería Biomédicainstname:Universidad de los Andesreponame:Repositorio Institucional SénecaFacial action unit detection with convolutional neural networksTrabajo de grado - Maestríainfo:eu-repo/semantics/masterThesishttp://purl.org/coar/version/c_970fb48d4fbd8a85Texthttp://purl.org/redcol/resource_type/TMRedes neurales (Computadores) - InvestigacionesProcesamiento de imágenes - InvestigacionesSistemas de reconocimiento de configuraciones - InvestigacionesExpresión facial - Procesamiento de imágenes - InvestigacionesIngenieríaPublicationTEXTu729464.pdf.txtu729464.pdf.txtExtracted texttext/plain50557https://repositorio.uniandes.edu.co/bitstreams/3224cc09-86be-4ffb-b704-7609c65e148f/downloadbb11deab464cc1168d5afd4215039fd1MD54THUMBNAILu729464.pdf.jpgu729464.pdf.jpgIM Thumbnailimage/jpeg29463https://repositorio.uniandes.edu.co/bitstreams/dec6061f-5519-4c1d-86c9-1fc491dc8e94/download8a55815adb40b48dc8e630f939b0aa4cMD55ORIGINALu729464.pdfapplication/pdf29945233https://repositorio.uniandes.edu.co/bitstreams/7cac9b95-df64-490c-8ab6-f0df701f0a44/downloadfc0e7b5aabc78210a1e61034f8d8ad26MD511992/13826oai:repositorio.uniandes.edu.co:1992/138262023-10-10 18:23:50.413http://creativecommons.org/licenses/by-nc-nd/4.0/open.accesshttps://repositorio.uniandes.edu.coRepositorio institucional Sénecaadminrepositorio@uniandes.edu.co
dc.title.es_CO.fl_str_mv Facial action unit detection with convolutional neural networks
title Facial action unit detection with convolutional neural networks
spellingShingle Facial action unit detection with convolutional neural networks
Redes neurales (Computadores) - Investigaciones
Procesamiento de imágenes - Investigaciones
Sistemas de reconocimiento de configuraciones - Investigaciones
Expresión facial - Procesamiento de imágenes - Investigaciones
Ingeniería
title_short Facial action unit detection with convolutional neural networks
title_full Facial action unit detection with convolutional neural networks
title_fullStr Facial action unit detection with convolutional neural networks
title_full_unstemmed Facial action unit detection with convolutional neural networks
title_sort Facial action unit detection with convolutional neural networks
dc.creator.fl_str_mv Romero Vergara, Andrés Felipe
dc.contributor.advisor.none.fl_str_mv Arbeláez Escalante, Pablo Andrés
dc.contributor.author.none.fl_str_mv Romero Vergara, Andrés Felipe
dc.subject.keyword.es_CO.fl_str_mv Redes neurales (Computadores) - Investigaciones
Procesamiento de imágenes - Investigaciones
Sistemas de reconocimiento de configuraciones - Investigaciones
Expresión facial - Procesamiento de imágenes - Investigaciones
topic Redes neurales (Computadores) - Investigaciones
Procesamiento de imágenes - Investigaciones
Sistemas de reconocimiento de configuraciones - Investigaciones
Expresión facial - Procesamiento de imágenes - Investigaciones
Ingeniería
dc.subject.themes.none.fl_str_mv Ingeniería
description We propose a novel deep convolutional neural network architecture to study the problem of action unit detection. We leverage recent gains in large-scale object recognition by formulating the task of predicting the presence of a specific action unit in a still image as simple image-level binary classification. We first train a convolutional encoder on the problem of multi-view emotion recognition as a high-level representation of facial expressions. We show that our architecture generalizes across views, ethnicity, gender and age by merging and training jointly on three standard emotion recognition datasets: CK+, Bosphorus and RafD. Our system is the first fully multi-view emotion recognizer proposed in the literature. We then extend this shared learned representation with fully-connected layers trained to detect individual action units. Our approach is conceptually simpler and yet significantly more accurate than the best methods based on the dominant paradigm for the study of this problem, which relies on facial landmark detection as an intermediate task. We conduct experiments on the BP4D dataset, the largest and most challenging benchmark currently available for action unit detection, and report an absolute improvement of 16% over the previous state-of-the-art.
publishDate 2017
dc.date.issued.es_CO.fl_str_mv 2017
dc.date.accessioned.none.fl_str_mv 2018-09-28T10:56:43Z
dc.date.available.none.fl_str_mv 2018-09-28T10:56:43Z
dc.type.spa.fl_str_mv Trabajo de grado - Maestría
dc.type.coarversion.fl_str_mv http://purl.org/coar/version/c_970fb48d4fbd8a85
dc.type.driver.spa.fl_str_mv info:eu-repo/semantics/masterThesis
dc.type.content.spa.fl_str_mv Text
dc.type.redcol.spa.fl_str_mv http://purl.org/redcol/resource_type/TM
dc.identifier.uri.none.fl_str_mv http://hdl.handle.net/1992/13826
dc.identifier.pdf.none.fl_str_mv u729464.pdf
dc.identifier.instname.spa.fl_str_mv instname:Universidad de los Andes
dc.identifier.reponame.spa.fl_str_mv reponame:Repositorio Institucional Séneca
dc.identifier.repourl.spa.fl_str_mv repourl:https://repositorio.uniandes.edu.co/
url http://hdl.handle.net/1992/13826
identifier_str_mv u729464.pdf
instname:Universidad de los Andes
reponame:Repositorio Institucional Séneca
repourl:https://repositorio.uniandes.edu.co/
dc.language.iso.es_CO.fl_str_mv eng
language eng
dc.rights.uri.*.fl_str_mv http://creativecommons.org/licenses/by-nc-nd/4.0/
dc.rights.accessrights.spa.fl_str_mv info:eu-repo/semantics/openAccess
dc.rights.coar.spa.fl_str_mv http://purl.org/coar/access_right/c_abf2
rights_invalid_str_mv http://creativecommons.org/licenses/by-nc-nd/4.0/
http://purl.org/coar/access_right/c_abf2
eu_rights_str_mv openAccess
dc.format.extent.es_CO.fl_str_mv 19 hojas
dc.format.mimetype.es_CO.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Uniandes
dc.publisher.program.es_CO.fl_str_mv Maestría en Ingeniería Biomédica
dc.publisher.faculty.es_CO.fl_str_mv Facultad de Ingeniería
dc.publisher.department.es_CO.fl_str_mv Departamento de Ingeniería Biomédica
publisher.none.fl_str_mv Uniandes
dc.source.es_CO.fl_str_mv instname:Universidad de los Andes
reponame:Repositorio Institucional Séneca
instname_str Universidad de los Andes
institution Universidad de los Andes
reponame_str Repositorio Institucional Séneca
collection Repositorio Institucional Séneca
bitstream.url.fl_str_mv https://repositorio.uniandes.edu.co/bitstreams/3224cc09-86be-4ffb-b704-7609c65e148f/download
https://repositorio.uniandes.edu.co/bitstreams/dec6061f-5519-4c1d-86c9-1fc491dc8e94/download
https://repositorio.uniandes.edu.co/bitstreams/7cac9b95-df64-490c-8ab6-f0df701f0a44/download
bitstream.checksum.fl_str_mv bb11deab464cc1168d5afd4215039fd1
8a55815adb40b48dc8e630f939b0aa4c
fc0e7b5aabc78210a1e61034f8d8ad26
bitstream.checksumAlgorithm.fl_str_mv MD5
MD5
MD5
repository.name.fl_str_mv Repositorio institucional Séneca
repository.mail.fl_str_mv adminrepositorio@uniandes.edu.co
_version_ 1812133992651030528