Person Re-Identification using Kernel-based Metric Learning Methods

Fei Xiong, Mengran Gou, Octavia Camps, Mario Sznaier
Northeastern Univeristy

[paper] [poster] [supp] [code]


Re-identification of individuals across camera networks with limited or no overlapping fields of view remains challenging in spite of significant research efforts. In this paper, we propose the use, and extensively evaluate the performance, of four alternatives for re-ID classification: regularized Pairwise Constrained Component Analysis, kernel Local Fisher Discriminant Analysis, Marginal Fisher Analysis and a ranking ensemble voting scheme, used in conjunction with different sizes of sets of histogram-based features and linear, X2 and RBF-X2 kernels. Comparisons against the state-of-art show significant improvements in performance measured both in terms of Cumulative Match Characteristic curves (CMC) and Proportion of Uncertainty Removed (PUR) scores on the challenging VIPeR, iLIDS, CAVIAR and 3DPeS datasets.

New! Collection of most public re-id datasets
New! The latest version of code is available on github now!


  • Very different features are used to evaluate the performance in the literature as shown below. We make a fair and comprehensive comparison with the state-of-art methods on four datasets and four different features.
  • We propose four kernel-based distance learning approaches to improve re-ID classification accuracy.
    Grid schema 8X16 patches, 8X8 stride 6 non-overlap strips 8X8 patches, 4X4 stride Hierarchical 6 strips, 32X32, 16X16 or 8X8 patches
    Feature channel HSV, Lab, LBP HSV, RGB, YUV, LBP HSV, YUV SIFT RGB, HSV, YUV, LBP
    Dimension reduction PCA N/A PCA Supervised PCA N/A
    Rank 1 results on VIPeR 19.60% 19.27% 24.18% 30.00% 35.10%

    Metric Learning Methods in Re-ID


    This work forcus on approaches designing classifiers to learn specialized metrics, that enforce features from the same individual to be closer than features from different individuals.

    Performances for Each Dataset

    VIPeRrestult iLIDSRrestult CAVIARrestult 3DPeSrestult

    Click for larger images. Click here to download data of above plots.

    Improving Performance by Fusing Algorithms



    X. Fei, M. Gou, O. Camps and M. Sznaier: "Person Re-Identification using Kernel-based Metric Learning Methods". In ECCV 2014.

        title={Person Re-Identification Using Kernel-Based Metric Learning Methods},
        author={Xiong, Fei and Gou, Mengran and Camps, Octavia and Sznaier, Mario},
        booktitle={Computer Vision--ECCV 2014},
    Hi, you are the -th visitor since 2015-09-8.