Transferring GANs

Transferring knowledge of pre-trained networks to new domains by means of fine-tuning is a widely used practice for applications based on discriminative models. To the best of our knowledge this practice has not been studied within the context of generative deep networks. Therefore, we study domain...

ver descrição completa

Detalhes bibliográficos
Autores: Wang, Yaxing|||0000-0002-6055-7164, Wu, Chenshen, Herranz, Luis|||0000-0002-7022-3395, Weijer, Joost van de|||0000-0002-9656-9706, Gonzalez-Garcia, Abel, Raducanu, Bogdan|||0000-0003-3648-8020
Tipo de documento: capítulo de livro
Data de publicação:2018
País:España
Recursos:Universitat Autònoma de Barcelona
Repositório:Dipòsit Digital de Documents de la UAB
Idioma:inglês
OAI Identifier:oai:ddd.uab.cat:304012
Acesso em linha:https://ddd.uab.cat/record/304012
https://dx.doi.org/urn:doi:10.1007/978-3-030-01231-1_14
Access Level:Acceso aberto
Palavra-chave:Generative adversarial networks
Transfer learning
Domain adaptation
Image generation
id ES_ce5a1ddb9dd8eb63f2cb6427e4f8daf7
oai_identifier_str oai:ddd.uab.cat:304012
network_acronym_str ES
network_name_str España
repository_id_str
spelling Transferring GANsGenerating images from limited dataWang, Yaxing|||0000-0002-6055-7164Wu, ChenshenHerranz, Luis|||0000-0002-7022-3395Weijer, Joost van de|||0000-0002-9656-9706Gonzalez-Garcia, AbelRaducanu, Bogdan|||0000-0003-3648-8020Generative adversarial networksTransfer learningDomain adaptationImage generationTransferring knowledge of pre-trained networks to new domains by means of fine-tuning is a widely used practice for applications based on discriminative models. To the best of our knowledge this practice has not been studied within the context of generative deep networks. Therefore, we study domain adaptation applied to image generation with generative adversarial networks. We evaluate several aspects of domain adaptation, including the impact of target domain size, the relative distance between source and target domain, and the initialization of conditional GANs. Our results show that using knowledge from pre-trained networks can shorten the convergence time and can significantly improve the quality of the generated images, especially when target data is limited. We show that these conclusions can also be drawn for conditional GANs even when the pre-trained model was trained without conditioning. Our results also suggest that density is more important than diversity and a dataset with one or few densely sampled classes is a better source model than more diverse datasets such as ImageNet or Places.Springer 22018-01-0120182018-01-01Capítol de llibrehttp://purl.org/coar/resource_type/c_3248AMhttp://purl.org/coar/version/c_ab4af688f83e57aainfo:eu-repo/semantics/bookPartapplication/pdfhttps://ddd.uab.cat/record/304012https://dx.doi.org/urn:doi:10.1007/978-3-030-01231-1_14reponame:Dipòsit Digital de Documents de la UABinstname:Universitat Autònoma de BarcelonaInglésengEuropean Commission https://doi.org/10.13039/501100000780 665919Agencia Estatal de Investigación https://doi.org/10.13039/501100011033 TIN2016-79717-RMinisterio de Economía y Competitividad https://doi.org/10.13039/501100003329 PCIN-2015-251open accesshttp://purl.org/coar/access_right/c_abf2Aquest material està protegit per drets d'autor i/o drets afins. Podeu utilitzar aquest material en funció del que permet la legislació de drets d'autor i drets afins d'aplicació al vostre cas. Per a d'altres usos heu d'obtenir permís del(s) titular(s) de drets.https://rightsstatements.org/vocab/InC/1.0/info:eu-repo/semantics/openAccessoai:ddd.uab.cat:3040122026-06-06T12:50:31Z
dc.title.none.fl_str_mv Transferring GANs
Generating images from limited data
title Transferring GANs
spellingShingle Transferring GANs
Wang, Yaxing|||0000-0002-6055-7164
Generative adversarial networks
Transfer learning
Domain adaptation
Image generation
title_short Transferring GANs
title_full Transferring GANs
title_fullStr Transferring GANs
title_full_unstemmed Transferring GANs
title_sort Transferring GANs
dc.creator.none.fl_str_mv Wang, Yaxing|||0000-0002-6055-7164
Wu, Chenshen
Herranz, Luis|||0000-0002-7022-3395
Weijer, Joost van de|||0000-0002-9656-9706
Gonzalez-Garcia, Abel
Raducanu, Bogdan|||0000-0003-3648-8020
author Wang, Yaxing|||0000-0002-6055-7164
author_facet Wang, Yaxing|||0000-0002-6055-7164
Wu, Chenshen
Herranz, Luis|||0000-0002-7022-3395
Weijer, Joost van de|||0000-0002-9656-9706
Gonzalez-Garcia, Abel
Raducanu, Bogdan|||0000-0003-3648-8020
author_role author
author2 Wu, Chenshen
Herranz, Luis|||0000-0002-7022-3395
Weijer, Joost van de|||0000-0002-9656-9706
Gonzalez-Garcia, Abel
Raducanu, Bogdan|||0000-0003-3648-8020
author2_role author
author
author
author
author
dc.subject.none.fl_str_mv Generative adversarial networks
Transfer learning
Domain adaptation
Image generation
topic Generative adversarial networks
Transfer learning
Domain adaptation
Image generation
description Transferring knowledge of pre-trained networks to new domains by means of fine-tuning is a widely used practice for applications based on discriminative models. To the best of our knowledge this practice has not been studied within the context of generative deep networks. Therefore, we study domain adaptation applied to image generation with generative adversarial networks. We evaluate several aspects of domain adaptation, including the impact of target domain size, the relative distance between source and target domain, and the initialization of conditional GANs. Our results show that using knowledge from pre-trained networks can shorten the convergence time and can significantly improve the quality of the generated images, especially when target data is limited. We show that these conclusions can also be drawn for conditional GANs even when the pre-trained model was trained without conditioning. Our results also suggest that density is more important than diversity and a dataset with one or few densely sampled classes is a better source model than more diverse datasets such as ImageNet or Places.
publishDate 2018
dc.date.none.fl_str_mv 2
2018-01-01
2018
2018-01-01
dc.type.none.fl_str_mv Capítol de llibre
http://purl.org/coar/resource_type/c_3248
AM
http://purl.org/coar/version/c_ab4af688f83e57aa
dc.type.openaire.fl_str_mv info:eu-repo/semantics/bookPart
format bookPart
dc.identifier.none.fl_str_mv https://ddd.uab.cat/record/304012
https://dx.doi.org/urn:doi:10.1007/978-3-030-01231-1_14
url https://ddd.uab.cat/record/304012
https://dx.doi.org/urn:doi:10.1007/978-3-030-01231-1_14
dc.language.none.fl_str_mv Inglés
eng
language_invalid_str_mv Inglés
language eng
dc.relation.none.fl_str_mv European Commission https://doi.org/10.13039/501100000780 665919
Agencia Estatal de Investigación https://doi.org/10.13039/501100011033 TIN2016-79717-R
Ministerio de Economía y Competitividad https://doi.org/10.13039/501100003329 PCIN-2015-251
dc.rights.none.fl_str_mv open access
http://purl.org/coar/access_right/c_abf2
https://rightsstatements.org/vocab/InC/1.0/
dc.rights.openaire.fl_str_mv info:eu-repo/semantics/openAccess
rights_invalid_str_mv open access
http://purl.org/coar/access_right/c_abf2
https://rightsstatements.org/vocab/InC/1.0/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Springer
publisher.none.fl_str_mv Springer
dc.source.none.fl_str_mv reponame:Dipòsit Digital de Documents de la UAB
instname:Universitat Autònoma de Barcelona
instname_str Universitat Autònoma de Barcelona
reponame_str Dipòsit Digital de Documents de la UAB
collection Dipòsit Digital de Documents de la UAB
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869419983287615488
score 15,811543

Registros relacionados