首页 > 其他分享> > person reid评价指标CMC原理及代码解析

person reid评价指标CMC原理及代码解析

2019-05-13 09:42:13 作者：互联网

CMC全程是Cumulative Matching Characteristics, 是行人重识别问题中的经典评价指标。该曲线的横坐标为rank，纵坐标为识别率百分比。rank n表示识别结果相似性降序排列中前n个结果包含目标。识别率是rank n 的数目#(rank n)占总的query样本数的比例。如下图

CMC曲线图来源 https://blog.csdn.net/m0_37240250/article/details/79942939

代码解释

先贴代码

代码来源 [港中文Xiaogang Wang教授主页]

下载链接

function C = cmc(D, varargin)
% CMC computes the Cumulative Match Characteristic.
%   C = CMC(D) computes the CMC of given distance matrix D between each pair
%   of gallery and probe person. Both gallery and probe persons are
%   assumed to be unique, i.e., the i-th gallery person matches only to
%   the i-th probe person. D(i, j) should be the distance between
%   gallery instance i and probe instance j.
%   单个参数C = CMC(D)时表示gallery和probe中的identity都是独一无二的，
%   即gallery中第i个person只匹配probe set第i个person
%
%   C = CMC(D, G, P) computes the CMC with G and P are id labels of each
%   person in gallery and probe sets. D is M*N where M and N are the lengths of
%   vector G and P, respectively. This function will first randomly select
%   exactly one instance among for each gallery identity and then calculate
%   the CMC of this sub-matrix. This procedure will be repeated 100 times
%   and the final result is the mean of them.
%   三个参数C = CMC(D, G, P),G和P是gallery和probe中person的id label,
%   所以D的高=G的个数，D的宽=P的个数
%   该脚本先对galleray中的每个identity随机挑选一个instance(同一个identity可能有多个instance),
%   计算子矩阵的CMC,然后重复一百次,求均值

switch nargin 
% nargin 针对当前正在执行的函数，返回函数调用中给定函数输入参数的数目
% nargin(fun) 返回 fun 函数定义中出现的输入参数的数目
% 如果该函数定义中包含 varargin，那么 nargin 返回输入数目的负数
% 例如，如果 myFun 函数声明输入 a、b 和 varargin，那么 nargin('myFun') 返回
% -3,就可以确定varargin前有2个参数
    case 1
        assert(ismatrix(D));
        assert(size(D, 1) == size(D, 2));% 判断是矩阵且是方阵

    case 3
        G = varargin{1};
        P = varargin{2};
        assert(isvector(G));
        assert(isvector(P));
        assert(length(G) == size(D, 1));
        assert(length(P) == size(D, 2));

    otherwise
        error('Invalid inputs');
end

if nargin == 1
    C = cmc_core(D, 1:size(D, 1), 1:size(D, 1));
else
    gtypes = union(G, []);%求集合，防止重复label
    ngtypes = length(gtypes);
    ntimes = 100;

    C = zeros(ngtypes, ntimes);%宽是100，有100条曲线求均值
    for t = 1:ntimes
        subdist = zeros(ngtypes, size(D, 2));
        for i = 1:ngtypes
            j = find(G == gtypes(i));%寻找gtypes(i)相同的G中label
            k = j(randi(length(j)));% j可能有多个值，从中随机选一个
            subdist(i, :) = D(k, :);%选取D中对应的行向量（该Label到probe set中每个label的距离的向量）
        end
        C(:, t) = cmc_core(subdist, gtypes, P);
    end
    C = mean(C, 2);
end

end


function C = cmc_core(D, G, P)
m = size(D, 1);
n = size(D, 2);
[~, order] = sort(D);%二维矩阵时，对矩阵的每一列分别进行升序排序,即query中每个label在gallery中相似度的升序
match = (G(order) == repmat(P, [m, 1]));%(m*n==m*n),找矩阵中对应相同元素
%repmat(P, [m, 1])是以P为unit构建m(unit)X1(unit)的矩阵
% G(order)与order/D大小一致，即按照order里面每列的排序重新排列G
C = cumsum(sum(match, 2) / n);
%sum(match, 2)每行累加
%cumsum是将矩阵按列进行求和，每一项是前面之和（类似于积分曲线）
end