Fixed FAR Vote Fusion of regional Facial Classifiers

Lieuwe Jan Spreeuwers, Raymond N.J. Veldhuis, S. Sultanali, J. Diephuis

    Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

    22 Downloads (Pure)

    Abstract

    Holistic face recognition methods like PCA and LDA have the disadvantage that they are very sensitive to expression, hair and illumination variations. This is one of the main reasons they are no longer competitive in the major benchmarks like FRGC and FRVT. In this paper we present an LDA based approach that combines many overlapping regional classifiers (experts) using what we call a Fixed FAR Voting Fusion (FFVF) strategy. The combination by voting of regional classifiers means that if there are sufficient regional classifiers unaffected by the expression, illumination or hair variations, the fused classifier will still correctly recognise the face. The FFVF approach has two interesting properties: it allows robust fusion of dependent classifiers and it only requires a single parameter to be tuned to obtain weights for fusion of different classifiers. We show the potential of the FFVF of regional classifiers using the standard benchmarks experiments 1 and 4 on FRGCv2 data. The multiregion FFVF classifier has a FRR of 4% at FAR=0.1% for controlled and 38% for uncontrolled data compared to 7% and 56% for the best single region classifier.
    Original languageUndefined
    Title of host publicationBIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group
    EditorsC. Busch, A. Brömme
    Place of PublicationDarmstadt
    PublisherGesellschaft für Informatik
    Pages1-4
    Number of pages8
    ISBN (Print)978-3-88579-624-4
    Publication statusPublished - Sep 2014
    Event13th International Conference of the Biometrics Special Interest Group, BIOSIG 2014 - Darmstadt, Germany
    Duration: 10 Sep 201412 Sep 2014
    Conference number: 13

    Publication series

    Name
    PublisherGesellschaft für Informatik

    Conference

    Conference13th International Conference of the Biometrics Special Interest Group, BIOSIG 2014
    Abbreviated titleBIOSIG 2014
    CountryGermany
    CityDarmstadt
    Period10/09/1412/09/14

    Keywords

    • EWI-24960
    • SCS-Safety
    • IR-91652
    • METIS-305972
    • Face recognition fusion regional

    Cite this

    Spreeuwers, L. J., Veldhuis, R. N. J., Sultanali, S., & Diephuis, J. (2014). Fixed FAR Vote Fusion of regional Facial Classifiers. In C. Busch, & A. Brömme (Eds.), BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group (pp. 1-4). Darmstadt: Gesellschaft für Informatik.
    Spreeuwers, Lieuwe Jan ; Veldhuis, Raymond N.J. ; Sultanali, S. ; Diephuis, J. / Fixed FAR Vote Fusion of regional Facial Classifiers. BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group. editor / C. Busch ; A. Brömme. Darmstadt : Gesellschaft für Informatik, 2014. pp. 1-4
    @inproceedings{a821c1d43d5946b9a1a5fb234d1ebab7,
    title = "Fixed FAR Vote Fusion of regional Facial Classifiers",
    abstract = "Holistic face recognition methods like PCA and LDA have the disadvantage that they are very sensitive to expression, hair and illumination variations. This is one of the main reasons they are no longer competitive in the major benchmarks like FRGC and FRVT. In this paper we present an LDA based approach that combines many overlapping regional classifiers (experts) using what we call a Fixed FAR Voting Fusion (FFVF) strategy. The combination by voting of regional classifiers means that if there are sufficient regional classifiers unaffected by the expression, illumination or hair variations, the fused classifier will still correctly recognise the face. The FFVF approach has two interesting properties: it allows robust fusion of dependent classifiers and it only requires a single parameter to be tuned to obtain weights for fusion of different classifiers. We show the potential of the FFVF of regional classifiers using the standard benchmarks experiments 1 and 4 on FRGCv2 data. The multiregion FFVF classifier has a FRR of 4{\%} at FAR=0.1{\%} for controlled and 38{\%} for uncontrolled data compared to 7{\%} and 56{\%} for the best single region classifier.",
    keywords = "EWI-24960, SCS-Safety, IR-91652, METIS-305972, Face recognition fusion regional",
    author = "Spreeuwers, {Lieuwe Jan} and Veldhuis, {Raymond N.J.} and S. Sultanali and J. Diephuis",
    year = "2014",
    month = "9",
    language = "Undefined",
    isbn = "978-3-88579-624-4",
    publisher = "Gesellschaft f{\"u}r Informatik",
    pages = "1--4",
    editor = "C. Busch and A. Br{\"o}mme",
    booktitle = "BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group",
    address = "Germany",

    }

    Spreeuwers, LJ, Veldhuis, RNJ, Sultanali, S & Diephuis, J 2014, Fixed FAR Vote Fusion of regional Facial Classifiers. in C Busch & A Brömme (eds), BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group. Gesellschaft für Informatik, Darmstadt, pp. 1-4, 13th International Conference of the Biometrics Special Interest Group, BIOSIG 2014, Darmstadt, Germany, 10/09/14.

    Fixed FAR Vote Fusion of regional Facial Classifiers. / Spreeuwers, Lieuwe Jan; Veldhuis, Raymond N.J.; Sultanali, S.; Diephuis, J.

    BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group. ed. / C. Busch; A. Brömme. Darmstadt : Gesellschaft für Informatik, 2014. p. 1-4.

    Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

    TY - GEN

    T1 - Fixed FAR Vote Fusion of regional Facial Classifiers

    AU - Spreeuwers, Lieuwe Jan

    AU - Veldhuis, Raymond N.J.

    AU - Sultanali, S.

    AU - Diephuis, J.

    PY - 2014/9

    Y1 - 2014/9

    N2 - Holistic face recognition methods like PCA and LDA have the disadvantage that they are very sensitive to expression, hair and illumination variations. This is one of the main reasons they are no longer competitive in the major benchmarks like FRGC and FRVT. In this paper we present an LDA based approach that combines many overlapping regional classifiers (experts) using what we call a Fixed FAR Voting Fusion (FFVF) strategy. The combination by voting of regional classifiers means that if there are sufficient regional classifiers unaffected by the expression, illumination or hair variations, the fused classifier will still correctly recognise the face. The FFVF approach has two interesting properties: it allows robust fusion of dependent classifiers and it only requires a single parameter to be tuned to obtain weights for fusion of different classifiers. We show the potential of the FFVF of regional classifiers using the standard benchmarks experiments 1 and 4 on FRGCv2 data. The multiregion FFVF classifier has a FRR of 4% at FAR=0.1% for controlled and 38% for uncontrolled data compared to 7% and 56% for the best single region classifier.

    AB - Holistic face recognition methods like PCA and LDA have the disadvantage that they are very sensitive to expression, hair and illumination variations. This is one of the main reasons they are no longer competitive in the major benchmarks like FRGC and FRVT. In this paper we present an LDA based approach that combines many overlapping regional classifiers (experts) using what we call a Fixed FAR Voting Fusion (FFVF) strategy. The combination by voting of regional classifiers means that if there are sufficient regional classifiers unaffected by the expression, illumination or hair variations, the fused classifier will still correctly recognise the face. The FFVF approach has two interesting properties: it allows robust fusion of dependent classifiers and it only requires a single parameter to be tuned to obtain weights for fusion of different classifiers. We show the potential of the FFVF of regional classifiers using the standard benchmarks experiments 1 and 4 on FRGCv2 data. The multiregion FFVF classifier has a FRR of 4% at FAR=0.1% for controlled and 38% for uncontrolled data compared to 7% and 56% for the best single region classifier.

    KW - EWI-24960

    KW - SCS-Safety

    KW - IR-91652

    KW - METIS-305972

    KW - Face recognition fusion regional

    M3 - Conference contribution

    SN - 978-3-88579-624-4

    SP - 1

    EP - 4

    BT - BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group

    A2 - Busch, C.

    A2 - Brömme, A.

    PB - Gesellschaft für Informatik

    CY - Darmstadt

    ER -

    Spreeuwers LJ, Veldhuis RNJ, Sultanali S, Diephuis J. Fixed FAR Vote Fusion of regional Facial Classifiers. In Busch C, Brömme A, editors, BIOSIG 2014: Proceedings of the 13th International Conference of the Biometrics Special Interest Group. Darmstadt: Gesellschaft für Informatik. 2014. p. 1-4