Fisher Vector Based Metric Learning Transformation For Kinship Verification

    1. Kinship Verification Problem

    Kinship Verification is a challenging task wherein given a pair of images (a parent and child image), the task is to determine whether a kinship relation exists between the parent and child image
    It is challenging task as the correlated visual resemblance between parents and their offspring have to be captured.

    2. Architecture


    2.1 Metric Learning Objective and Algorithm

    Objective
    Minimize the distance if the parent and child image have kinship relation
    Maximize the distance if no kinship relation



    3. Experiments

    Datasets Used
    KinFaceW-I
    KinFaceW-II
    Four kinship relations
    father-son
    father-daughter
    mother-son
    mother-daughter


    Figure: Some example images from both datasets are shown below.



    Table: Comparison between the proposed method and other
    popular learning methods on the KinfaceW-I dataset.



    Table: Comparison between the proposed method and other
    popular learning methods on the KinfaceW-II dataset.

    References

    [1] CSML - Cosine Similarity Metric Learning
    HieuV. Nguyen and Li Bai, "Cosine similarity metric learning for face verification," in ACCV, 2011, vol. 6493, pp. 709-720.
    [2] NCA - Neighbourhood Components Analysis
    Jacob Goldberger, Sam T. Roweis, Geoffrey E. Hinton, and Ruslan Salakhutdinov, "Neighbourhood components analysis," in NIPS, 2004.
    [3] LMNN - Large Margin Nearest Neighbor Metric Learning
    Kilian Q. Weinberger and Lawrence K. Saul, "Distance metric learning for large margin nearest neighbor classification," J. Mach. Learn. Res., vol. 10, pp. 207-244, 2009.
    [4] NRML - Neighborhood Repulsed Metric Learning
    Jiwen Lu, Xiuzhuang Zhou, Yap-Pen Tan, Yuanyuan Shang, and Jie Zhou, "Neighborhood repulsed metric learning for kinship verification," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 36, no. 2, pp. 331-345, 2014.
    [5] MNRML - Multi-neighborhood Repulsed Metric Learning
    Jiwen Lu, Xiuzhuang Zhou, Yap-Pen Tan, Yuanyuan Shang, and Jie Zhou, "Neighborhood repulsed metric learning for kinship verification," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 36, no. 2, pp. 331-345, 2014.
    [6] ITML - Information-Theoretic Metric Learning
    Jason V. Davis, Brian Kulis, Prateek Jain, Suvrit Sra, and Inderjit S. Dhillon, "Information-theoretic metric learning," in ICML, 2007, pp. 209-216.
    [7] GGA - Generative Gated Autoencoders
    Afshin Dehghan, Enrique G Ortiz, Ruben Villegas, and Mubarak Shah, "Who do i look like? determining parent-offspring resemblance via gated autoencoders," in CVPR. IEEE, 2014, pp. 1757-1764.
    [8] DGA - Discriminative Gated Autoencoders
    Afshin Dehghan, Enrique G Ortiz, Ruben Villegas, and Mubarak Shah, "Who do i look like? determining parent-offspring resemblance via gated autoencoders," in CVPR. IEEE, 2014, pp. 1757-1764.