MNIST数据库介绍及转换

 MNIST数据库介绍：MNIST是一个手写数字数据库，它有60000个训练样本集和10000个测试样本集。它是NIST数据库的一个子集。

         MNIST数据库官方网址为：http://yann.lecun.com/exdb/mnist/ ，也可以在windows下直接下载，train-images-idx3-ubyte.gz、train-labels-idx1-ubyte.gz等。下载四个文件，解压缩。解压缩后发现这些文件并不是标准的图像格式。这些图像数据都保存在二进制文件中。每个样本图像的宽高为28*28。

         以下为将其转换成普通的jpg图像格式的代码：

#include "funset.hpp"#include <iostream>#include <fstream>#include <vector>#include <opencv2/opencv.hpp>static int ReverseInt(int i){ unsigned char ch1, ch2, ch3, ch4; ch1 = i & 255; ch2 = (i >> 8) & 255; ch3 = (i >> 16) & 255; ch4 = (i >> 24) & 255; return((int)ch1 << 24) + ((int)ch2 << 16) + ((int)ch3 << 8) + ch4;}static void read_Mnist(std::string filename, std::vector<cv::Mat> &vec){ std::ifstream file(filename, std::ios::binary); if (file.is_open()) {  int magic_number = 0;  int number_of_images = 0;  int n_rows = 0;  int n_cols = 0;  file.read((char*)&magic_number, sizeof(magic_number));  magic_number = ReverseInt(magic_number);  file.read((char*)&number_of_images, sizeof(number_of_images));  number_of_images = ReverseInt(number_of_images);  file.read((char*)&n_rows, sizeof(n_rows));  n_rows = ReverseInt(n_rows);  file.read((char*)&n_cols, sizeof(n_cols));  n_cols = ReverseInt(n_cols);  for (int i = 0; i < number_of_images; ++i) {   cv::Mat tp = cv::Mat::zeros(n_rows, n_cols, CV_8UC1);   for (int r = 0; r < n_rows; ++r) {    for (int c = 0; c < n_cols; ++c) {     unsigned char temp = 0;     file.read((char*)&temp, sizeof(temp));     tp.at<uchar>(r, c) = (int)temp;    }   }   vec.push_back(tp);  } }}static void read_Mnist_Label(std::string filename, std::vector<int> &vec){ std::ifstream file(filename, std::ios::binary); if (file.is_open()) {  int magic_number = 0;  int number_of_images = 0;  int n_rows = 0;  int n_cols = 0;  file.read((char*)&magic_number, sizeof(magic_number));  magic_number = ReverseInt(magic_number);  file.read((char*)&number_of_images, sizeof(number_of_images));  number_of_images = ReverseInt(number_of_images);  for (int i = 0; i < number_of_images; ++i) {   unsigned char temp = 0;   file.read((char*)&temp, sizeof(temp));   vec[i] = (int)temp;  } }}static std::string GetImageName(int number, int arr[]){ std::string str1, str2; for (int i = 0; i < 10; i++) {  if (number == i) {   arr[i]++;   str1 = std::to_string(arr[i]);   if (arr[i] < 10) {    str1 = "0000" + str1;   } else if (arr[i] < 100) {    str1 = "000" + str1;   } else if (arr[i] < 1000) {    str1 = "00" + str1;   } else if (arr[i] < 10000) {    str1 = "0" + str1;   }   break;  } } str2 = std::to_string(number) + "_" + str1; return str2;}int MNISTtoImage(){ // reference: http://eric-yuan.me/cpp-read-mnist/ // test images and test labels // read MNIST image into OpenCV Mat vector std::string filename_test_images = "E:/GitCode/NN_Test/data/database/MNIST/t10k-images.idx3-ubyte"; int number_of_test_images = 10000; std::vector<cv::Mat> vec_test_images; read_Mnist(filename_test_images, vec_test_images); // read MNIST label into int vector std::string filename_test_labels = "E:/GitCode/NN_Test/data/database/MNIST/t10k-labels.idx1-ubyte"; std::vector<int> vec_test_labels(number_of_test_images); read_Mnist_Label(filename_test_labels, vec_test_labels); if (vec_test_images.size() != vec_test_labels.size()) {  std::cout << "parse MNIST test file error" << std::endl;  return -1; } // save test images int count_digits[10]; std::fill(&count_digits[0], &count_digits[0] + 10, 0); std::string save_test_images_path = "E:/GitCode/NN_Test/data/tmp/MNIST/test_images/"; for (int i = 0; i < vec_test_images.size(); i++) {  int number = vec_test_labels[i];  std::string image_name = GetImageName(number, count_digits);  image_name = save_test_images_path + image_name + ".jpg";  cv::imwrite(image_name, vec_test_images[i]); } // train images and train labels // read MNIST image into OpenCV Mat vector std::string filename_train_images = "E:/GitCode/NN_Test/data/database/MNIST/train-images.idx3-ubyte"; int number_of_train_images = 60000; std::vector<cv::Mat> vec_train_images; read_Mnist(filename_train_images, vec_train_images); // read MNIST label into int vector std::string filename_train_labels = "E:/GitCode/NN_Test/data/database/MNIST/train-labels.idx1-ubyte"; std::vector<int> vec_train_labels(number_of_train_images); read_Mnist_Label(filename_train_labels, vec_train_labels); if (vec_train_images.size() != vec_train_labels.size()) {  std::cout << "parse MNIST train file error" << std::endl;  return -1; } // save train images std::fill(&count_digits[0], &count_digits[0] + 10, 0); std::string save_train_images_path = "E:/GitCode/NN_Test/data/tmp/MNIST/train_images/"; for (int i = 0; i < vec_train_images.size(); i++) {  int number = vec_train_labels[i];  std::string image_name = GetImageName(number, count_digits);  image_name = save_train_images_path + image_name + ".jpg";  cv::imwrite(image_name, vec_train_images[i]); } // save big imags std::string images_path = "E:/GitCode/NN_Test/data/tmp/MNIST/train_images/"; int width = 28 * 20; int height = 28 * 10; cv::Mat dst(height, width, CV_8UC1); for (int i = 0; i < 10; i++) {  for (int j = 1; j <= 20; j++) {   int x = (j-1) * 28;   int y = i * 28;   cv::Mat part = dst(cv::Rect(x, y, 28, 28));   std::string str = std::to_string(j);   if (j < 10)    str = "0000" + str;   else    str = "000" + str;   str = std::to_string(i) + "_" + str + ".jpg";   std::string input_image = images_path + str;   cv::Mat src = cv::imread(input_image, 0);   if (src.empty()) {    fprintf(stderr, "read image error: %s\n", input_image.c_str());    return -1;   }   src.copyTo(part);  } } std::string output_image = images_path + "result.png"; cv::imwrite(output_image, dst); return 0;}