Semantic Segmentation of Urban Scenes via Domain Adaptation of SYNTHIA

Vision-based semantic Semantic segmentation in urban scenarios is a key functionality for autonomous driving. Recent revolutionary results of deep convolutional neural networks (CNNs) foreshadow the advent of reliable classifiers to perform such visual tasks. However, CNNs require learning of many p...

Full description

Saved in:
Bibliographic Details
Published inDomain Adaptation in Computer Vision Applications pp. 227 - 241
Main Authors Ros, German, Sellart, Laura, Villalonga, Gabriel, Maidanik, Elias, Molero, Francisco, Garcia, Marc, Cedeño, Adriana, Perez, Francisco, Ramirez, Didier, Escobar, Eduardo, Gomez, Jose Luis, Vazquez, David, Lopez, Antonio M.
Format Book Chapter
LanguageEnglish
Published Switzerland Springer International Publishing AG 2017
Springer International Publishing
SeriesAdvances in Computer Vision and Pattern Recognition
Subjects
Online AccessGet full text
ISBN3319583468
9783319583464
ISSN2191-6586
2191-6594
DOI10.1007/978-3-319-58347-1_12

Cover

Loading…
Abstract Vision-based semantic Semantic segmentation in urban scenarios is a key functionality for autonomous driving. Recent revolutionary results of deep convolutional neural networks (CNNs) foreshadow the advent of reliable classifiers to perform such visual tasks. However, CNNs require learning of many parameters from raw images; thus, having a sufficient amount of diverse images with class annotations is needed. These annotations are obtained via cumbersome, human labor which is particularly challenging for semantic segmentation since pixel-level annotations are required. In this chapter, we propose to use a combination of a virtual Virtual world to automatically generate realistic synthetic images with pixel-level annotations, and domain adaptation to transfer the models learned to correctly operate in real scenarios. We address the question of how useful synthetic Synthetic data can be for semantic segmentation—Semantic segmentation particular, when using a CNN paradigm. In order to answer this question we have generated a synthetic collection of diverse urban images, named SYNTHIA, with automatically generated class annotations and object identifiers. We use SYNTHIA in combination with publicly available real-world urban images with manually provided annotations. Then, we conduct experiments with CNNs that show that combining SYNTHIA with simple domain adaptation techniques in the training stage significantly improves performance on semantic segmentation.
AbstractList Vision-based semantic Semantic segmentation in urban scenarios is a key functionality for autonomous driving. Recent revolutionary results of deep convolutional neural networks (CNNs) foreshadow the advent of reliable classifiers to perform such visual tasks. However, CNNs require learning of many parameters from raw images; thus, having a sufficient amount of diverse images with class annotations is needed. These annotations are obtained via cumbersome, human labor which is particularly challenging for semantic segmentation since pixel-level annotations are required. In this chapter, we propose to use a combination of a virtual Virtual world to automatically generate realistic synthetic images with pixel-level annotations, and domain adaptation to transfer the models learned to correctly operate in real scenarios. We address the question of how useful synthetic Synthetic data can be for semantic segmentation—Semantic segmentation particular, when using a CNN paradigm. In order to answer this question we have generated a synthetic collection of diverse urban images, named SYNTHIA, with automatically generated class annotations and object identifiers. We use SYNTHIA in combination with publicly available real-world urban images with manually provided annotations. Then, we conduct experiments with CNNs that show that combining SYNTHIA with simple domain adaptation techniques in the training stage significantly improves performance on semantic segmentation.
Author Villalonga, Gabriel
Molero, Francisco
Vazquez, David
Ramirez, Didier
Escobar, Eduardo
Perez, Francisco
Sellart, Laura
Ros, German
Garcia, Marc
Cedeño, Adriana
Gomez, Jose Luis
Maidanik, Elias
Lopez, Antonio M.
Author_xml – sequence: 1
  givenname: German
  surname: Ros
  fullname: Ros, German
  email: gros@cvc.uab.es
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 2
  givenname: Laura
  surname: Sellart
  fullname: Sellart, Laura
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 3
  givenname: Gabriel
  surname: Villalonga
  fullname: Villalonga, Gabriel
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 4
  givenname: Elias
  surname: Maidanik
  fullname: Maidanik, Elias
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 5
  givenname: Francisco
  surname: Molero
  fullname: Molero, Francisco
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 6
  givenname: Marc
  surname: Garcia
  fullname: Garcia, Marc
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 7
  givenname: Adriana
  surname: Cedeño
  fullname: Cedeño, Adriana
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 8
  givenname: Francisco
  surname: Perez
  fullname: Perez, Francisco
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 9
  givenname: Didier
  surname: Ramirez
  fullname: Ramirez, Didier
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 10
  givenname: Eduardo
  surname: Escobar
  fullname: Escobar, Eduardo
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 11
  givenname: Jose Luis
  surname: Gomez
  fullname: Gomez, Jose Luis
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 12
  givenname: David
  surname: Vazquez
  fullname: Vazquez, David
  organization: Computer Vision Center, Barcelona, Spain
– sequence: 13
  givenname: Antonio M.
  surname: Lopez
  fullname: Lopez, Antonio M.
  organization: Computer Vision Center, Barcelona, Spain
BookMark eNpFkM9SwjAQxqOiIyBv4KEvEM0m26Q5MijCDKOHwsFTJi0pViHFpvr8BvHPZXfm-_bb2f0NSM833hFyDewGGFO3WmVUUAGapplARcEAPyGjKIsofmtwSvocNFCZajwjg19DZr0_I5MXZBAXokAtAS_JKIRXxhhkKsVU9Mk0dzvru7pMcrfZOd_Zrm580lTJqi2sT_LSeReSz9omd83O1j4Zr-3-fyp_flzO5uMrcl7ZbXCjnz4kq-n9cjKji6eH-WS8oHuOoqOlUghaaC0LXqpYUTi7zhjTFmW1xsJBqQoRL6tYUWClnBClQm5RpbYCKYaEH_eGfVv7jWtN0TRvwQAzB2wm8jHCRBDmG5E5YIshPIb2bfP-4UJn3CEVP-tauy1f4j-uDSZlmEIqDAAazqX4Au38a5k
ContentType Book Chapter
Copyright Springer International Publishing AG 2017
Copyright_xml – notice: Springer International Publishing AG 2017
DBID FFUUA
DEWEY 006.37
DOI 10.1007/978-3-319-58347-1_12
DatabaseName ProQuest Ebook Central - Book Chapters - Demo use only
DatabaseTitleList
DeliveryMethod fulltext_linktorsrc
Discipline Applied Sciences
Computer Science
EISBN 9783319583471
3319583476
EISSN 2191-6594
Editor Csurka, Gabriela
Editor_xml – sequence: 1
  fullname: Csurka, Gabriela
EndPage 241
ExternalDocumentID EBC5045153_114_226
GroupedDBID 0D6
0DA
20A
38.
AABBV
AALVI
ABHTH
ABQUB
ACBPT
ACDJR
ADCXD
AEJLV
AEKFX
AETDV
AEZAY
AGIGN
AGYGE
AIODD
ALBAV
ALMA_UNASSIGNED_HOLDINGS
AZZ
BATQV
BBABE
CRSEL
CZZ
FFUUA
I4C
IEZ
JJU
SBO
SWYDZ
TPJZQ
Z5O
Z7R
Z7S
Z7U
Z7W
Z7X
Z7Y
Z7Z
Z81
Z82
Z83
Z84
Z85
Z87
Z88
ID FETCH-LOGICAL-p243t-c774193996b2c796b43ead8009a46fd4be1c7b3453f0bb4f7e33c742a475af163
ISBN 3319583468
9783319583464
ISSN 2191-6586
IngestDate Tue Jul 29 20:02:39 EDT 2025
Mon Apr 07 01:52:08 EDT 2025
IsPeerReviewed false
IsScholarly false
LCCallNum TA1634
Language English
LinkModel OpenURL
MergedId FETCHMERGED-LOGICAL-p243t-c774193996b2c796b43ead8009a46fd4be1c7b3453f0bb4f7e33c742a475af163
OCLC 1004349614
PQID EBC5045153_114_226
PageCount 15
ParticipantIDs springer_books_10_1007_978_3_319_58347_1_12
proquest_ebookcentralchapters_5045153_114_226
PublicationCentury 2000
PublicationDate 2017
PublicationDateYYYYMMDD 2017-01-01
PublicationDate_xml – year: 2017
  text: 2017
PublicationDecade 2010
PublicationPlace Switzerland
PublicationPlace_xml – name: Switzerland
– name: Cham
PublicationSeriesTitle Advances in Computer Vision and Pattern Recognition
PublicationSeriesTitleAlternate Advs Comp. Vision, Pattern Recognition
PublicationTitle Domain Adaptation in Computer Vision Applications
PublicationYear 2017
Publisher Springer International Publishing AG
Springer International Publishing
Publisher_xml – name: Springer International Publishing AG
– name: Springer International Publishing
RelatedPersons Kang, Sing Bing
RelatedPersons_xml – sequence: 1
  givenname: Sing Bing
  surname: Kang
  fullname: Kang, Sing Bing
  organization: Microsoft Research, Microsoft (United States), Redmond, USA
SSID ssj0001875453
ssib023166760
ssib006652368
Score 1.5195348
Snippet Vision-based semantic Semantic segmentation in urban scenarios is a key functionality for autonomous driving. Recent revolutionary results of deep...
SourceID springer
proquest
SourceType Publisher
StartPage 227
SubjectTerms Artificial intelligence
Image processing
Public administration
Title Semantic Segmentation of Urban Scenes via Domain Adaptation of SYNTHIA
URI http://ebookcentral.proquest.com/lib/SITE_ID/reader.action?docID=5045153&ppg=226
http://link.springer.com/10.1007/978-3-319-58347-1_12
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV09b9swECVcdyk69BtNv8Chm8FCEinRHjIERVI3KLLENtKJICWyMJrIQeR06C_Lz-sdSVmy4yVdBEMgZPF4ON47vXck5HMipUnHpmAl7HVM6FyySVE5ZnSVV2Jcudx6lu9ZMZ2L04v8YjC467GWbtfmS_l3r67kf1YV7sG6okr2ASu7eSjcgN-wvnCFFYbrTvK7XWYNXJbVlcZiRaWvI2PQC_jCIQ2jhdeMhySzX5ULhymGajjqBTYVFmRBBfkOSqU3wXqBpxJdrupQf_2mDUDrjkqol5Wul78jPywSjaIDnlt4PLaDPbe_rqLCyeem8xvjgwpG2dGfpR7dnwkqaX6ezabfg3-hMW1zeBT4Cs2-iXrRg-8Vip8kIikqulwsaqRyp6jRFjV3yqJdZW4LBXOOHXO4CO3QQ_CEQJwyyK5im-3-vYnoB-zQmSDu_VlownVvW-kzSVD1hf8mWarweOtHcpwPyeOj49Mfiy6SFQDwOyAHSTRyiZOu-gcwUfgWqZsXReFRO5FxaA3VTawn-tz3FlvwaOeLvk-UZs_JUxTPUFS1gElfkIGtX5JnEerQuJE0r8hJ6x207x105aj3Dhq8g4J30OAdtPMOHBW94zWZnxzPvk5ZPNKDXWeCr1kJaAMgA4Bsk5USroJDKAPQMtGicJUwNi2l4WAalxgjnLScl1JkWshcO8AOb8iwXtX2LaGZs0lWwqBsYoWrLDwBwIFOHOTY2A__gLDWJsoTDyLbuQwWaFSOrZVyjrp8BSDkgIxawykc3qi2ozdYXHEFFlfe4got_u5Bo9-TJ52XfyDD9c2t_QjJ7Np8ip7zD83elaY
linkProvider Library Specific Holdings
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=bookitem&rft.title=Domain+Adaptation+in+Computer+Vision+Applications&rft.au=Ros%2C+German&rft.au=Sellart%2C+Laura&rft.au=Villalonga%2C+Gabriel&rft.au=Maidanik%2C+Elias&rft.atitle=Semantic+Segmentation+of+Urban+Scenes+via+Domain+Adaptation+of+SYNTHIA&rft.series=Advances+in+Computer+Vision+and+Pattern+Recognition&rft.date=2017-01-01&rft.pub=Springer+International+Publishing&rft.isbn=9783319583464&rft.issn=2191-6586&rft.eissn=2191-6594&rft.spage=227&rft.epage=241&rft_id=info:doi/10.1007%2F978-3-319-58347-1_12
thumbnail_s http://utb.summon.serialssolutions.com/2.0.0/image/custom?url=https%3A%2F%2Febookcentral.proquest.com%2Fcovers%2F5045153-l.jpg