Learning-based Face Synthesis for Pose-Robust Recognition from Single Image

Akshay Asthana, Conrad Sanderson, Tom Gedeon, Roland Goecke

    Research output: A Conference proceeding or a Chapter in BookConference contribution

    30 Citations (Scopus)

    Abstract

    Face recognition in real-world conditions requires the ability to deal with a number
    of conditions, such as variations in pose, illumination and expression. In this paper, we
    focus on variations in head pose and use a computationally efficient regression-based
    approach for synthesising face images in different poses, which are used to extend the
    face recognition training set. In this data-driven approach, the correspondences between
    facial landmark points in frontal and non-frontal views are learnt offline from manually
    annotated training data via Gaussian Process Regression. We then use this learner to
    synthesise non-frontal face images from any unseen frontal image. To demonstrate the
    utility of this approach, two frontal face recognition systems (the commonly used PCA
    and the recent Multi-Region Histograms) are augmented with synthesised non-frontal
    views for each person. This synthesis and augmentation approach is experimentally
    validated on the FERET dataset, showing a considerable improvement in recognition
    rates for ±40◦ and ±60◦ views, while maintaining high recognition rates for ±15◦
    and ±25◦ views.
    Original languageEnglish
    Title of host publicationBritish Machine Vision Conference 2009
    EditorsAndrea Cavallaro, Simon Prince, Daniel Alexander
    Place of PublicationLondon, UK
    PublisherBritish Machine Vision Association
    Pages1-10
    Number of pages10
    DOIs
    Publication statusPublished - 2009
    EventBritish Machine Vision Conference 2009 - London, United Kingdom
    Duration: 7 Sep 200910 Sep 2009

    Conference

    ConferenceBritish Machine Vision Conference 2009
    CountryUnited Kingdom
    CityLondon
    Period7/09/0910/09/09

    Fingerprint

    Face recognition
    Lighting

    Cite this

    Asthana, A., Sanderson, C., Gedeon, T., & Goecke, R. (2009). Learning-based Face Synthesis for Pose-Robust Recognition from Single Image. In A. Cavallaro, S. Prince, & D. Alexander (Eds.), British Machine Vision Conference 2009 (pp. 1-10). London, UK: British Machine Vision Association. https://doi.org/10.5244/C.23.31
    Asthana, Akshay ; Sanderson, Conrad ; Gedeon, Tom ; Goecke, Roland. / Learning-based Face Synthesis for Pose-Robust Recognition from Single Image. British Machine Vision Conference 2009. editor / Andrea Cavallaro ; Simon Prince ; Daniel Alexander. London, UK : British Machine Vision Association, 2009. pp. 1-10
    @inproceedings{defec55aa74d4b0ea4bc1ffc463f7ea6,
    title = "Learning-based Face Synthesis for Pose-Robust Recognition from Single Image",
    abstract = "Face recognition in real-world conditions requires the ability to deal with a numberof conditions, such as variations in pose, illumination and expression. In this paper, wefocus on variations in head pose and use a computationally efficient regression-basedapproach for synthesising face images in different poses, which are used to extend theface recognition training set. In this data-driven approach, the correspondences betweenfacial landmark points in frontal and non-frontal views are learnt offline from manuallyannotated training data via Gaussian Process Regression. We then use this learner tosynthesise non-frontal face images from any unseen frontal image. To demonstrate theutility of this approach, two frontal face recognition systems (the commonly used PCAand the recent Multi-Region Histograms) are augmented with synthesised non-frontalviews for each person. This synthesis and augmentation approach is experimentallyvalidated on the FERET dataset, showing a considerable improvement in recognitionrates for ±40◦ and ±60◦ views, while maintaining high recognition rates for ±15◦and ±25◦ views.",
    author = "Akshay Asthana and Conrad Sanderson and Tom Gedeon and Roland Goecke",
    year = "2009",
    doi = "10.5244/C.23.31",
    language = "English",
    pages = "1--10",
    editor = "Andrea Cavallaro and Simon Prince and Daniel Alexander",
    booktitle = "British Machine Vision Conference 2009",
    publisher = "British Machine Vision Association",

    }

    Asthana, A, Sanderson, C, Gedeon, T & Goecke, R 2009, Learning-based Face Synthesis for Pose-Robust Recognition from Single Image. in A Cavallaro, S Prince & D Alexander (eds), British Machine Vision Conference 2009. British Machine Vision Association, London, UK, pp. 1-10, British Machine Vision Conference 2009, London, United Kingdom, 7/09/09. https://doi.org/10.5244/C.23.31

    Learning-based Face Synthesis for Pose-Robust Recognition from Single Image. / Asthana, Akshay; Sanderson, Conrad; Gedeon, Tom; Goecke, Roland.

    British Machine Vision Conference 2009. ed. / Andrea Cavallaro; Simon Prince; Daniel Alexander. London, UK : British Machine Vision Association, 2009. p. 1-10.

    Research output: A Conference proceeding or a Chapter in BookConference contribution

    TY - GEN

    T1 - Learning-based Face Synthesis for Pose-Robust Recognition from Single Image

    AU - Asthana, Akshay

    AU - Sanderson, Conrad

    AU - Gedeon, Tom

    AU - Goecke, Roland

    PY - 2009

    Y1 - 2009

    N2 - Face recognition in real-world conditions requires the ability to deal with a numberof conditions, such as variations in pose, illumination and expression. In this paper, wefocus on variations in head pose and use a computationally efficient regression-basedapproach for synthesising face images in different poses, which are used to extend theface recognition training set. In this data-driven approach, the correspondences betweenfacial landmark points in frontal and non-frontal views are learnt offline from manuallyannotated training data via Gaussian Process Regression. We then use this learner tosynthesise non-frontal face images from any unseen frontal image. To demonstrate theutility of this approach, two frontal face recognition systems (the commonly used PCAand the recent Multi-Region Histograms) are augmented with synthesised non-frontalviews for each person. This synthesis and augmentation approach is experimentallyvalidated on the FERET dataset, showing a considerable improvement in recognitionrates for ±40◦ and ±60◦ views, while maintaining high recognition rates for ±15◦and ±25◦ views.

    AB - Face recognition in real-world conditions requires the ability to deal with a numberof conditions, such as variations in pose, illumination and expression. In this paper, wefocus on variations in head pose and use a computationally efficient regression-basedapproach for synthesising face images in different poses, which are used to extend theface recognition training set. In this data-driven approach, the correspondences betweenfacial landmark points in frontal and non-frontal views are learnt offline from manuallyannotated training data via Gaussian Process Regression. We then use this learner tosynthesise non-frontal face images from any unseen frontal image. To demonstrate theutility of this approach, two frontal face recognition systems (the commonly used PCAand the recent Multi-Region Histograms) are augmented with synthesised non-frontalviews for each person. This synthesis and augmentation approach is experimentallyvalidated on the FERET dataset, showing a considerable improvement in recognitionrates for ±40◦ and ±60◦ views, while maintaining high recognition rates for ±15◦and ±25◦ views.

    U2 - 10.5244/C.23.31

    DO - 10.5244/C.23.31

    M3 - Conference contribution

    SP - 1

    EP - 10

    BT - British Machine Vision Conference 2009

    A2 - Cavallaro, Andrea

    A2 - Prince, Simon

    A2 - Alexander, Daniel

    PB - British Machine Vision Association

    CY - London, UK

    ER -

    Asthana A, Sanderson C, Gedeon T, Goecke R. Learning-based Face Synthesis for Pose-Robust Recognition from Single Image. In Cavallaro A, Prince S, Alexander D, editors, British Machine Vision Conference 2009. London, UK: British Machine Vision Association. 2009. p. 1-10 https://doi.org/10.5244/C.23.31