最近做图像检索的内容，要用到cifar-10，网上大都很乱。因此自己写了一个简单的解析cifar-10 的matlab代码。
你需到官网下载matlab版本数据。
%解析cifar-10数据matlab代码 cifar网址：http://www.cs.toronto.edu/~kriz/cifar.html %这只是一个简单的数据解析代码，你首先应设置一个全局变量 number=[1,...1];10个1 %由于时间紧迫，这个代码并没有考虑简洁性问题，但是更加容易理解。 % createcifar.m % @param phase = ‘train’ or ‘test’ % @param number is a vector ; length = 10 ;records number of each class image imwrited function [number] = createcifar(phase,number,data,labels,resize_method) for index = 1:10000 % all matlab loop variables change from 1 not 0! % include channels of an image pixelr = data(index,1:1024); pixelg = data(index,1025:2048); pixelb = data(index,2049:3072); class = labels(index); for line = 1:32 image(line,:,1)=(pixelr(1,(line-1)*32+1:line*32)); image(line,:,2)=(pixelg(1,(line-1)*32+1:line*32)); image(line,:,3)=(pixelb(1,(line-1)*32+1:line*32)); end result = imresize(image,[224,224],resize_method); switch(class) case 0 %note this part can be replaced a variable 'PATH' imwrite(result,char(['E:\',phase,'\airplane\',num2str(number(1)),'.jpg'])); number(1) = number(1)+1; case 1 imwrite(result,char(['E:\',phase,'\automobile\',num2str(number(2)),'.jpg'])); number(2) = number(2)+1; case 2 imwrite(result,char(['E:\',phase,'\bird\',num2str(number(3)),'.jpg'])); number(3) = number(3)+1; case 3 imwrite(result,char(['E:\',phase,'\cat\',num2str(number(4)),'.jpg'])); number(4) = number(4)+1; case 4 imwrite(result,char(['E:\',phase,'\deer\',num2str(number(5)),'.jpg'])); number(5) = number(5)+1; case 5 imwrite(result,char(['E:\',phase,'\dog\',num2str(number(6)),'.jpg'])); number(6) = number(6)+1; case 6 imwrite(result,char(['E:\',phase,'\frog\',num2str(number(7)),'.jpg'])); number(7) = number(7)+1; case 7 imwrite(result,char(['E:\',phase,'\horse\',num2str(number(8)),'.jpg'])); number(8) = number(8)+1; case 8 imwrite(result,char(['E:\',phase,'\ship\',num2str(number(9)),'.jpg'])); number(9) = number(9)+1; case 9 imwrite(result,char(['E:\',phase,'\truck\',num2str(number(10)),'.jpg'])); number(10) = number(10)+1; end end end
% main function : main.m clc; clear all; number=[1,1,1,1,1,1,1,1,1,1]; phase = 'train' method = 'bicubic' % 'bilinear' ,'bicubic','nearest' for batchi = 1:5 load(['D:\Download.bak\Downloads\cifar-10-matlab\cifar-10-batches-mat\data_batch_',num2str(batchi),'.mat']); number = createcifar(phase,number,data,labels,method); end number=[1,1,1,1,1,1,1,1,1,1]; phase = 'test' load('D:\Download.bak\Downloads\cifar-10-matlab\cifar-10-batches-mat\test_batch.mat'); number = createcifar(phase,number,data,labels,method);
cifar-10 地址：http://pan.baidu.com/s/1cF2gEy