论文PCANet: A Simple Deep Learning Baseline for Image Classification?的matlab源码解读（四）

PCANet_train.m函数

<span style="font-family:Times New Roman;font-size:18px;"><span style="font-size:18px;">function [f V BlkIdx] = PCANet_train(InImg,PCANet,IdtExt)
% =======INPUT=============
% InImg     Input images (cell); each cell can be either a matrix (Gray) or a 3D tensor (RGB)  
% PCANet    PCANet parameters (struct)
%       .PCANet.NumStages      
%           the number of stages in PCANet; e.g., 2  
%       .PatchSize
%           the patch size (filter size) for square patches; e.g., [5 3]  %patchsize是方阵
%           means patch size equalt to 5 and 3 in the first stage and second stage, respectively 
%       .NumFilters
%           the number of filters in each stage; e.g., [16 8] means 16 and
%           8 filters in the first stage and second stage, respectively
%       .HistBlockSize 
%           the size of each block for local histogram; e.g., [10 10]
%       .BlkOverLapRatio 重叠块区域比
%           overlapped block region ratio; e.g., 0 means no overlapped 
%           between blocks, and 0.3 means 30% of blocksize is overlapped 
%       .Pyramid 空间金字塔匹配
%           spatial pyramid matching; e.g., [1 2 4], and [] if no Pyramid
%           is applied
% IdtExt    a number in {0,1}; 1 do feature extraction, and 0 otherwise  </span>
<span style="font-size:18px;">% =======OUTPUT============
% f         PCANet features (each column corresponds to feature of each image)一列对应一副图像的特征
% V         learned PCA filter banks (cell)滤波器组
% BlkIdx    index of local block from which the histogram is compuated 局部块索引 
% =========================

addpath('./Utils')%将上一级目录Utils添加到matlab的工作路径中，Utils：实用工具 


if length(PCANet.NumFilters)~= PCANet.NumStages;  %~=是不等于
    display('Length(PCANet.NumFilters)~=PCANet.NumStages')  %显示矩阵和文字内容的重载函数
    return
end

NumImg = length(InImg);

V = cell(PCANet.NumStages,1); %创建每一阶段的PCA滤波器组
OutImg = InImg; 
ImgIdx = (1:NumImg)';% 1到NumImg的整数，注意是一个列向量
clear InImg; 

for stage = 1:PCANet.NumStages
    display(['Computing PCA filter bank and its outputs at stage ' num2str(stage) '...'])%num2str()把数值转换为字符串
    </span>
   <span style="font-size:18px;"> V{stage} = PCA_FilterBank(OutImg, PCANet.PatchSize(stage), PCANet.NumFilters(stage)); % compute PCA filter banks
    
    if stage ~= PCANet.NumStages % compute the PCA outputs only when it is NOT the last stage
        [OutImg ImgIdx] = PCA_output(OutImg, ImgIdx, ...
            PCANet.PatchSize(stage), PCANet.NumFilters(stage), V{stage});  
    end
    % 注意经过此次循环，ImgIdx不再是(1:NumImg)'，被PCA_output.m赋值了！！
end

if IdtExt == 1 % enable feature extraction
    %display('PCANet training feature extraction...') 
    
    f = cell(NumImg,1); % compute the PCANet training feature one by one 
    
    for idx = 1:NumImg
        if 0==mod(idx,100); display(['Extracting PCANet feasture of the ' num2str(idx) 'th training sample...']); end
        OutImgIndex = ImgIdx==idx; % select feature maps corresponding to image "idx" (outputs of the-last-but-one PCA filter bank) 
        %这里的ImgIdx是一个列向量，idx是一个数字，此句返回ImgIdx的列向量中值=idx的位置值设为1，其余值设为0
        [OutImg_i ImgIdx_i] = PCA_output(OutImg(OutImgIndex), ones(sum(OutImgIndex),1),...
            PCANet.PatchSize(end), PCANet.NumFilters(end), V{end});  % compute the last PCA outputs of image "idx"
        %sum(OutImgIndex)返回矩阵中1的个数，这里就是数值大小应该 等于L1*L2
        [f{idx} BlkIdx] = HashingHist(PCANet,ImgIdx_i,OutImg_i); % compute the feature of image "idx"
%        [f{idx} BlkIdx] = SphereSum(PCANet,ImgIdx_i,OutImg_i); % Testing!!
        OutImg(OutImgIndex) = cell(sum(OutImgIndex),1); 
       </span>
   <span style="font-size:18px;"> end
    f = sparse([f{:}]);
    
else  % disable feature extraction
    f = [];
    BlkIdx = [];
end</span></span>