Semantic pyramids for gender and action recognition.
Identifieur interne : 001056 ( Main/Exploration ); précédent : 001055; suivant : 001057Semantic pyramids for gender and action recognition.
Auteurs : Fahad Shahbaz Khan ; Joost Van De Weijer ; Rao Muhammad Anwer ; Michael Felsberg ; Carlo GattaSource :
- IEEE transactions on image processing : a publication of the IEEE Signal Processing Society [ 1941-0042 ] ; 2014.
Descripteurs français
- KwdFr :
- Actigraphie (méthodes), Algorithmes (MeSH), Amélioration d'image (méthodes), Biométrie (méthodes), Détermination du sexe (méthodes), Femelle (MeSH), Humains (MeSH), Imagerie du corps entier (méthodes), Intelligence artificielle (MeSH), Interprétation d'images assistée par ordinateur (méthodes), Mâle (MeSH), Reconnaissance automatique des formes (méthodes), Reproductibilité des résultats (MeSH), Sensibilité et spécificité (MeSH), Sémantique (MeSH).
- MESH :
- méthodes : Actigraphie, Amélioration d'image, Biométrie, Détermination du sexe, Imagerie du corps entier, Interprétation d'images assistée par ordinateur, Reconnaissance automatique des formes.
- Algorithmes, Femelle, Humains, Intelligence artificielle, Mâle, Reproductibilité des résultats, Sensibilité et spécificité, Sémantique.
English descriptors
- KwdEn :
- Actigraphy (methods), Algorithms (MeSH), Artificial Intelligence (MeSH), Biometry (methods), Female (MeSH), Humans (MeSH), Image Enhancement (methods), Image Interpretation, Computer-Assisted (methods), Male (MeSH), Pattern Recognition, Automated (methods), Reproducibility of Results (MeSH), Semantics (MeSH), Sensitivity and Specificity (MeSH), Sex Determination Analysis (methods), Whole Body Imaging (methods).
- MESH :
Abstract
Person description is a challenging problem in computer vision. We investigated two major aspects of person description: 1) gender and 2) action recognition in still images. Most state-of-the-art approaches for gender and action recognition rely on the description of a single body part, such as face or full-body. However, relying on a single body part is suboptimal due to significant variations in scale, viewpoint, and pose in real-world images. This paper proposes a semantic pyramid approach for pose normalization. Our approach is fully automatic and based on combining information from full-body, upper-body, and face regions for gender and action recognition in still images. The proposed approach does not require any annotations for upper-body and face of a person. Instead, we rely on pretrained state-of-the-art upper-body and face detectors to automatically extract semantic information of a person. Given multiple bounding boxes from each body part detector, we then propose a simple method to select the best candidate bounding box, which is used for feature extraction. Finally, the extracted features from the full-body, upper-body, and face regions are combined into a single representation for classification. To validate the proposed approach for gender recognition, experiments are performed on three large data sets namely: 1) human attribute; 2) head-shoulder; and 3) proxemics. For action recognition, we perform experiments on four data sets most used for benchmarking action recognition in still images: 1) Sports; 2) Willow; 3) PASCAL VOC 2010; and 4) Stanford-40. Our experiments clearly demonstrate that the proposed approach, despite its simplicity, outperforms state-of-the-art methods for gender and action recognition.
DOI: 10.1109/TIP.2014.2331759
PubMed: 24956369
Affiliations:
Links toward previous steps (curation, corpus...)
Le document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en">Semantic pyramids for gender and action recognition.</title>
<author><name sortKey="Khan, Fahad Shahbaz" sort="Khan, Fahad Shahbaz" uniqKey="Khan F" first="Fahad Shahbaz" last="Khan">Fahad Shahbaz Khan</name>
</author>
<author><name sortKey="Van De Weijer, Joost" sort="Van De Weijer, Joost" uniqKey="Van De Weijer J" first="Joost" last="Van De Weijer">Joost Van De Weijer</name>
</author>
<author><name sortKey="Anwer, Rao Muhammad" sort="Anwer, Rao Muhammad" uniqKey="Anwer R" first="Rao Muhammad" last="Anwer">Rao Muhammad Anwer</name>
</author>
<author><name sortKey="Felsberg, Michael" sort="Felsberg, Michael" uniqKey="Felsberg M" first="Michael" last="Felsberg">Michael Felsberg</name>
</author>
<author><name sortKey="Gatta, Carlo" sort="Gatta, Carlo" uniqKey="Gatta C" first="Carlo" last="Gatta">Carlo Gatta</name>
</author>
</titleStmt>
<publicationStmt><idno type="wicri:source">PubMed</idno>
<date when="2014">2014</date>
<idno type="RBID">pubmed:24956369</idno>
<idno type="pmid">24956369</idno>
<idno type="doi">10.1109/TIP.2014.2331759</idno>
<idno type="wicri:Area/Main/Corpus">001087</idno>
<idno type="wicri:explorRef" wicri:stream="Main" wicri:step="Corpus" wicri:corpus="PubMed">001087</idno>
<idno type="wicri:Area/Main/Curation">001087</idno>
<idno type="wicri:explorRef" wicri:stream="Main" wicri:step="Curation">001087</idno>
<idno type="wicri:Area/Main/Exploration">001087</idno>
</publicationStmt>
<sourceDesc><biblStruct><analytic><title xml:lang="en">Semantic pyramids for gender and action recognition.</title>
<author><name sortKey="Khan, Fahad Shahbaz" sort="Khan, Fahad Shahbaz" uniqKey="Khan F" first="Fahad Shahbaz" last="Khan">Fahad Shahbaz Khan</name>
</author>
<author><name sortKey="Van De Weijer, Joost" sort="Van De Weijer, Joost" uniqKey="Van De Weijer J" first="Joost" last="Van De Weijer">Joost Van De Weijer</name>
</author>
<author><name sortKey="Anwer, Rao Muhammad" sort="Anwer, Rao Muhammad" uniqKey="Anwer R" first="Rao Muhammad" last="Anwer">Rao Muhammad Anwer</name>
</author>
<author><name sortKey="Felsberg, Michael" sort="Felsberg, Michael" uniqKey="Felsberg M" first="Michael" last="Felsberg">Michael Felsberg</name>
</author>
<author><name sortKey="Gatta, Carlo" sort="Gatta, Carlo" uniqKey="Gatta C" first="Carlo" last="Gatta">Carlo Gatta</name>
</author>
</analytic>
<series><title level="j">IEEE transactions on image processing : a publication of the IEEE Signal Processing Society</title>
<idno type="eISSN">1941-0042</idno>
<imprint><date when="2014" type="published">2014</date>
</imprint>
</series>
</biblStruct>
</sourceDesc>
</fileDesc>
<profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>Actigraphy (methods)</term>
<term>Algorithms (MeSH)</term>
<term>Artificial Intelligence (MeSH)</term>
<term>Biometry (methods)</term>
<term>Female (MeSH)</term>
<term>Humans (MeSH)</term>
<term>Image Enhancement (methods)</term>
<term>Image Interpretation, Computer-Assisted (methods)</term>
<term>Male (MeSH)</term>
<term>Pattern Recognition, Automated (methods)</term>
<term>Reproducibility of Results (MeSH)</term>
<term>Semantics (MeSH)</term>
<term>Sensitivity and Specificity (MeSH)</term>
<term>Sex Determination Analysis (methods)</term>
<term>Whole Body Imaging (methods)</term>
</keywords>
<keywords scheme="KwdFr" xml:lang="fr"><term>Actigraphie (méthodes)</term>
<term>Algorithmes (MeSH)</term>
<term>Amélioration d'image (méthodes)</term>
<term>Biométrie (méthodes)</term>
<term>Détermination du sexe (méthodes)</term>
<term>Femelle (MeSH)</term>
<term>Humains (MeSH)</term>
<term>Imagerie du corps entier (méthodes)</term>
<term>Intelligence artificielle (MeSH)</term>
<term>Interprétation d'images assistée par ordinateur (méthodes)</term>
<term>Mâle (MeSH)</term>
<term>Reconnaissance automatique des formes (méthodes)</term>
<term>Reproductibilité des résultats (MeSH)</term>
<term>Sensibilité et spécificité (MeSH)</term>
<term>Sémantique (MeSH)</term>
</keywords>
<keywords scheme="MESH" qualifier="methods" xml:lang="en"><term>Actigraphy</term>
<term>Biometry</term>
<term>Image Enhancement</term>
<term>Image Interpretation, Computer-Assisted</term>
<term>Pattern Recognition, Automated</term>
<term>Sex Determination Analysis</term>
<term>Whole Body Imaging</term>
</keywords>
<keywords scheme="MESH" qualifier="méthodes" xml:lang="fr"><term>Actigraphie</term>
<term>Amélioration d'image</term>
<term>Biométrie</term>
<term>Détermination du sexe</term>
<term>Imagerie du corps entier</term>
<term>Interprétation d'images assistée par ordinateur</term>
<term>Reconnaissance automatique des formes</term>
</keywords>
<keywords scheme="MESH" xml:lang="en"><term>Algorithms</term>
<term>Artificial Intelligence</term>
<term>Female</term>
<term>Humans</term>
<term>Male</term>
<term>Reproducibility of Results</term>
<term>Semantics</term>
<term>Sensitivity and Specificity</term>
</keywords>
<keywords scheme="MESH" xml:lang="fr"><term>Algorithmes</term>
<term>Femelle</term>
<term>Humains</term>
<term>Intelligence artificielle</term>
<term>Mâle</term>
<term>Reproductibilité des résultats</term>
<term>Sensibilité et spécificité</term>
<term>Sémantique</term>
</keywords>
</textClass>
</profileDesc>
</teiHeader>
<front><div type="abstract" xml:lang="en">Person description is a challenging problem in computer vision. We investigated two major aspects of person description: 1) gender and 2) action recognition in still images. Most state-of-the-art approaches for gender and action recognition rely on the description of a single body part, such as face or full-body. However, relying on a single body part is suboptimal due to significant variations in scale, viewpoint, and pose in real-world images. This paper proposes a semantic pyramid approach for pose normalization. Our approach is fully automatic and based on combining information from full-body, upper-body, and face regions for gender and action recognition in still images. The proposed approach does not require any annotations for upper-body and face of a person. Instead, we rely on pretrained state-of-the-art upper-body and face detectors to automatically extract semantic information of a person. Given multiple bounding boxes from each body part detector, we then propose a simple method to select the best candidate bounding box, which is used for feature extraction. Finally, the extracted features from the full-body, upper-body, and face regions are combined into a single representation for classification. To validate the proposed approach for gender recognition, experiments are performed on three large data sets namely: 1) human attribute; 2) head-shoulder; and 3) proxemics. For action recognition, we perform experiments on four data sets most used for benchmarking action recognition in still images: 1) Sports; 2) Willow; 3) PASCAL VOC 2010; and 4) Stanford-40. Our experiments clearly demonstrate that the proposed approach, despite its simplicity, outperforms state-of-the-art methods for gender and action recognition. </div>
</front>
</TEI>
<pubmed><MedlineCitation Status="MEDLINE" Owner="NLM"><PMID Version="1">24956369</PMID>
<DateCompleted><Year>2015</Year>
<Month>09</Month>
<Day>29</Day>
</DateCompleted>
<DateRevised><Year>2014</Year>
<Month>08</Month>
<Day>15</Day>
</DateRevised>
<Article PubModel="Print-Electronic"><Journal><ISSN IssnType="Electronic">1941-0042</ISSN>
<JournalIssue CitedMedium="Internet"><Volume>23</Volume>
<Issue>8</Issue>
<PubDate><Year>2014</Year>
<Month>Aug</Month>
</PubDate>
</JournalIssue>
<Title>IEEE transactions on image processing : a publication of the IEEE Signal Processing Society</Title>
<ISOAbbreviation>IEEE Trans Image Process</ISOAbbreviation>
</Journal>
<ArticleTitle>Semantic pyramids for gender and action recognition.</ArticleTitle>
<Pagination><MedlinePgn>3633-45</MedlinePgn>
</Pagination>
<ELocationID EIdType="doi" ValidYN="Y">10.1109/TIP.2014.2331759</ELocationID>
<Abstract><AbstractText>Person description is a challenging problem in computer vision. We investigated two major aspects of person description: 1) gender and 2) action recognition in still images. Most state-of-the-art approaches for gender and action recognition rely on the description of a single body part, such as face or full-body. However, relying on a single body part is suboptimal due to significant variations in scale, viewpoint, and pose in real-world images. This paper proposes a semantic pyramid approach for pose normalization. Our approach is fully automatic and based on combining information from full-body, upper-body, and face regions for gender and action recognition in still images. The proposed approach does not require any annotations for upper-body and face of a person. Instead, we rely on pretrained state-of-the-art upper-body and face detectors to automatically extract semantic information of a person. Given multiple bounding boxes from each body part detector, we then propose a simple method to select the best candidate bounding box, which is used for feature extraction. Finally, the extracted features from the full-body, upper-body, and face regions are combined into a single representation for classification. To validate the proposed approach for gender recognition, experiments are performed on three large data sets namely: 1) human attribute; 2) head-shoulder; and 3) proxemics. For action recognition, we perform experiments on four data sets most used for benchmarking action recognition in still images: 1) Sports; 2) Willow; 3) PASCAL VOC 2010; and 4) Stanford-40. Our experiments clearly demonstrate that the proposed approach, despite its simplicity, outperforms state-of-the-art methods for gender and action recognition. </AbstractText>
</Abstract>
<AuthorList CompleteYN="Y"><Author ValidYN="Y"><LastName>Khan</LastName>
<ForeName>Fahad Shahbaz</ForeName>
<Initials>FS</Initials>
</Author>
<Author ValidYN="Y"><LastName>van de Weijer</LastName>
<ForeName>Joost</ForeName>
<Initials>J</Initials>
</Author>
<Author ValidYN="Y"><LastName>Anwer</LastName>
<ForeName>Rao Muhammad</ForeName>
<Initials>RM</Initials>
</Author>
<Author ValidYN="Y"><LastName>Felsberg</LastName>
<ForeName>Michael</ForeName>
<Initials>M</Initials>
</Author>
<Author ValidYN="Y"><LastName>Gatta</LastName>
<ForeName>Carlo</ForeName>
<Initials>C</Initials>
</Author>
</AuthorList>
<Language>eng</Language>
<PublicationTypeList><PublicationType UI="D016428">Journal Article</PublicationType>
<PublicationType UI="D013485">Research Support, Non-U.S. Gov't</PublicationType>
</PublicationTypeList>
<ArticleDate DateType="Electronic"><Year>2014</Year>
<Month>06</Month>
<Day>18</Day>
</ArticleDate>
</Article>
<MedlineJournalInfo><Country>United States</Country>
<MedlineTA>IEEE Trans Image Process</MedlineTA>
<NlmUniqueID>9886191</NlmUniqueID>
<ISSNLinking>1057-7149</ISSNLinking>
</MedlineJournalInfo>
<CitationSubset>IM</CitationSubset>
<MeshHeadingList><MeshHeading><DescriptorName UI="D056044" MajorTopicYN="N">Actigraphy</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="Y">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D000465" MajorTopicYN="N">Algorithms</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D001185" MajorTopicYN="N">Artificial Intelligence</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D001699" MajorTopicYN="N">Biometry</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="Y">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D005260" MajorTopicYN="N">Female</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D006801" MajorTopicYN="N">Humans</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D007089" MajorTopicYN="N">Image Enhancement</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="N">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D007090" MajorTopicYN="N">Image Interpretation, Computer-Assisted</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="Y">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D008297" MajorTopicYN="N">Male</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D010363" MajorTopicYN="N">Pattern Recognition, Automated</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="Y">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D015203" MajorTopicYN="N">Reproducibility of Results</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D012660" MajorTopicYN="N">Semantics</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D012680" MajorTopicYN="N">Sensitivity and Specificity</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D012732" MajorTopicYN="N">Sex Determination Analysis</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="Y">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName UI="D051598" MajorTopicYN="N">Whole Body Imaging</DescriptorName>
<QualifierName UI="Q000379" MajorTopicYN="Y">methods</QualifierName>
</MeshHeading>
</MeshHeadingList>
</MedlineCitation>
<PubmedData><History><PubMedPubDate PubStatus="entrez"><Year>2014</Year>
<Month>6</Month>
<Day>24</Day>
<Hour>6</Hour>
<Minute>0</Minute>
</PubMedPubDate>
<PubMedPubDate PubStatus="pubmed"><Year>2014</Year>
<Month>6</Month>
<Day>24</Day>
<Hour>6</Hour>
<Minute>0</Minute>
</PubMedPubDate>
<PubMedPubDate PubStatus="medline"><Year>2015</Year>
<Month>9</Month>
<Day>30</Day>
<Hour>6</Hour>
<Minute>0</Minute>
</PubMedPubDate>
</History>
<PublicationStatus>ppublish</PublicationStatus>
<ArticleIdList><ArticleId IdType="pubmed">24956369</ArticleId>
<ArticleId IdType="doi">10.1109/TIP.2014.2331759</ArticleId>
</ArticleIdList>
</PubmedData>
</pubmed>
<affiliations><list></list>
<tree><noCountry><name sortKey="Anwer, Rao Muhammad" sort="Anwer, Rao Muhammad" uniqKey="Anwer R" first="Rao Muhammad" last="Anwer">Rao Muhammad Anwer</name>
<name sortKey="Felsberg, Michael" sort="Felsberg, Michael" uniqKey="Felsberg M" first="Michael" last="Felsberg">Michael Felsberg</name>
<name sortKey="Gatta, Carlo" sort="Gatta, Carlo" uniqKey="Gatta C" first="Carlo" last="Gatta">Carlo Gatta</name>
<name sortKey="Khan, Fahad Shahbaz" sort="Khan, Fahad Shahbaz" uniqKey="Khan F" first="Fahad Shahbaz" last="Khan">Fahad Shahbaz Khan</name>
<name sortKey="Van De Weijer, Joost" sort="Van De Weijer, Joost" uniqKey="Van De Weijer J" first="Joost" last="Van De Weijer">Joost Van De Weijer</name>
</noCountry>
</tree>
</affiliations>
</record>
Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Bois/explor/WillowV1/Data/Main/Exploration
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 001056 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 001056 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien |wiki= Bois |area= WillowV1 |flux= Main |étape= Exploration |type= RBID |clé= pubmed:24956369 |texte= Semantic pyramids for gender and action recognition. }}
Pour générer des pages wiki
HfdIndexSelect -h $EXPLOR_AREA/Data/Main/Exploration/RBID.i -Sk "pubmed:24956369" \ | HfdSelect -Kh $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd \ | NlmPubMed2Wicri -a WillowV1
![]() | This area was generated with Dilib version V0.6.37. | ![]() |