Learning AAM fitting through simulation

Jason Saragih, Roland Goecke

    Research output: Contribution to journalArticle

    56 Citations (Scopus)

    Abstract

    The active appearance model (AAM) is a powerful method for modeling and segmenting deformable visual
    objects. The utility of the AAM stems from two fronts: its compact representation as a linear object class
    and its rapid fitting procedure, which utilizes fixed linear updates. Although the original fitting procedure
    works well for objects with restricted variability when initialization is close to the optimum, its efficacy
    deteriorates in more general settings, with regards to both accuracy and capture range. In this paper, we
    propose a novel fitting procedure where training is coupled with, and directly addresses, AAM fitting in
    its deployment. This is achieved by simulating the conditions of real fitting problems and learning the
    best set of fixed linear mappings, such that performance over these simulations is optimized. The power
    of the approach does not stem from an update model with larger capacity, but from addressing the whole
    fitting procedure simultaneously. To motivate the approach, it is compared with a number of existing
    AAM fitting procedures on two publicly available face databases. It is shown that this method exhibits
    convergence rates, capture range and convergence accuracy that are significantly better than other linear
    methods and comparable to a nonlinear method, whilst affording superior computational efficiency
    Original languageEnglish
    Pages (from-to)2628-2636
    Number of pages9
    JournalPattern Recognition
    Volume42
    Issue number11
    DOIs
    Publication statusPublished - 2009

    Fingerprint

    Computational efficiency
    Problem-Based Learning

    Cite this

    Saragih, Jason ; Goecke, Roland. / Learning AAM fitting through simulation. In: Pattern Recognition. 2009 ; Vol. 42, No. 11. pp. 2628-2636.
    @article{895340b3e5e34bcea5df1264f86c844f,
    title = "Learning AAM fitting through simulation",
    abstract = "The active appearance model (AAM) is a powerful method for modeling and segmenting deformable visualobjects. The utility of the AAM stems from two fronts: its compact representation as a linear object classand its rapid fitting procedure, which utilizes fixed linear updates. Although the original fitting procedureworks well for objects with restricted variability when initialization is close to the optimum, its efficacydeteriorates in more general settings, with regards to both accuracy and capture range. In this paper, wepropose a novel fitting procedure where training is coupled with, and directly addresses, AAM fitting inits deployment. This is achieved by simulating the conditions of real fitting problems and learning thebest set of fixed linear mappings, such that performance over these simulations is optimized. The powerof the approach does not stem from an update model with larger capacity, but from addressing the wholefitting procedure simultaneously. To motivate the approach, it is compared with a number of existingAAM fitting procedures on two publicly available face databases. It is shown that this method exhibitsconvergence rates, capture range and convergence accuracy that are significantly better than other linearmethods and comparable to a nonlinear method, whilst affording superior computational efficiency",
    keywords = "Active appearance model, Fitting, Discriminative, Linear model",
    author = "Jason Saragih and Roland Goecke",
    year = "2009",
    doi = "10.1016/j.patcog.2009.04.014",
    language = "English",
    volume = "42",
    pages = "2628--2636",
    journal = "Pattern Recognition",
    issn = "0031-3203",
    publisher = "Elsevier Limited",
    number = "11",

    }

    Learning AAM fitting through simulation. / Saragih, Jason; Goecke, Roland.

    In: Pattern Recognition, Vol. 42, No. 11, 2009, p. 2628-2636.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - Learning AAM fitting through simulation

    AU - Saragih, Jason

    AU - Goecke, Roland

    PY - 2009

    Y1 - 2009

    N2 - The active appearance model (AAM) is a powerful method for modeling and segmenting deformable visualobjects. The utility of the AAM stems from two fronts: its compact representation as a linear object classand its rapid fitting procedure, which utilizes fixed linear updates. Although the original fitting procedureworks well for objects with restricted variability when initialization is close to the optimum, its efficacydeteriorates in more general settings, with regards to both accuracy and capture range. In this paper, wepropose a novel fitting procedure where training is coupled with, and directly addresses, AAM fitting inits deployment. This is achieved by simulating the conditions of real fitting problems and learning thebest set of fixed linear mappings, such that performance over these simulations is optimized. The powerof the approach does not stem from an update model with larger capacity, but from addressing the wholefitting procedure simultaneously. To motivate the approach, it is compared with a number of existingAAM fitting procedures on two publicly available face databases. It is shown that this method exhibitsconvergence rates, capture range and convergence accuracy that are significantly better than other linearmethods and comparable to a nonlinear method, whilst affording superior computational efficiency

    AB - The active appearance model (AAM) is a powerful method for modeling and segmenting deformable visualobjects. The utility of the AAM stems from two fronts: its compact representation as a linear object classand its rapid fitting procedure, which utilizes fixed linear updates. Although the original fitting procedureworks well for objects with restricted variability when initialization is close to the optimum, its efficacydeteriorates in more general settings, with regards to both accuracy and capture range. In this paper, wepropose a novel fitting procedure where training is coupled with, and directly addresses, AAM fitting inits deployment. This is achieved by simulating the conditions of real fitting problems and learning thebest set of fixed linear mappings, such that performance over these simulations is optimized. The powerof the approach does not stem from an update model with larger capacity, but from addressing the wholefitting procedure simultaneously. To motivate the approach, it is compared with a number of existingAAM fitting procedures on two publicly available face databases. It is shown that this method exhibitsconvergence rates, capture range and convergence accuracy that are significantly better than other linearmethods and comparable to a nonlinear method, whilst affording superior computational efficiency

    KW - Active appearance model

    KW - Fitting

    KW - Discriminative

    KW - Linear model

    U2 - 10.1016/j.patcog.2009.04.014

    DO - 10.1016/j.patcog.2009.04.014

    M3 - Article

    VL - 42

    SP - 2628

    EP - 2636

    JO - Pattern Recognition

    JF - Pattern Recognition

    SN - 0031-3203

    IS - 11

    ER -