from tensorflow.contrib.learn.python.learn.datasets.mnist import read_data_sets

mnist = read_data_sets("data", one_hot=True, reshape=False, validation_size=0)

 import numpy, imageio, glob, sys, os, random

 #Imageio 提供简单的用于读写图像数据的接口

 #glob 功能类似于文件搜索，查找文件只用到三个匹配符：”*”, “?”, “[]”。”*”匹配0个或多个字符；”?”匹配单个字符；”[]”匹配指定范围内的字符，如：[0-9]匹配数字。

 def get_labels_and_files(folder, number):

   # Make a list of lists of files for each label

   filelists = []

   for label in range(0,10):

     filelist = []

     filelists.append(filelist);

     dirname = os.path.join(folder, chr(ord('A') + label))

     #label实际为0-9，chr(ord('A') + label)返回A-J

     #拼接路径dirname=folder/[A-J]

     for file in os.listdir(dirname):

     #返回一个装满当前路径中文件名的list

       if (file.endswith('.png')):

         fullname = os.path.join(dirname, file)

         if (os.path.getsize(fullname) > 0):

           filelist.append(fullname)

         else:

           print('file ' + fullname + ' is empty')

     # sort each list of files so they start off in the same order

     # regardless of how the order the OS returns them in

     filelist.sort()

   # Take the specified number of items for each label and

   # build them into an array of (label, filename) pairs

   # Since we seeded the RNG, we should get the same sample each run

   labelsAndFiles = []

   for label in range(0,10):

     filelist = random.sample(filelists[label], number)

     #随机采样 设定个数的文件名

     for filename in filelist:

       labelsAndFiles.append((label, filename))

       #Python的元组与列表类似，不同之处在于元组的元素不能修改。元组使用小括号，列表使用方括号。

   return labelsAndFiles

 def make_arrays(labelsAndFiles):

   images = []

   labels = []

   for i in range(0, len(labelsAndFiles)):

     # display progress, since this can take a while

     if (i % 100 == 0):

       sys.stdout.write("\r%d%% complete" % ((i * 100)/len(labelsAndFiles)))

       #\r 返回第一个指针，覆盖前面的内容

       sys.stdout.flush()

     filename = labelsAndFiles[i][1]

     try:

       image = imageio.imread(filename)

       images.append(image)

       labels.append(labelsAndFiles[i][0])

     except:

       # If this happens we won't have the requested number

       print("\nCan't read image file " + filename)

   count = len(images)

   imagedata = numpy.zeros((count,28,28), dtype=numpy.uint8)

   labeldata = numpy.zeros(count, dtype=numpy.uint8)

   for i in range(0, len(labelsAndFiles)):

     imagedata[i] = images[i]

     labeldata[i] = labels[i]

   print("\n")

   return imagedata, labeldata

 def write_labeldata(labeldata, outputfile):

   header = numpy.array([0x0801, len(labeldata)], dtype='>i4')

   with open(outputfile, "wb") as f:

   #以二进制写模式打开

   #这里使用了 with 语句，不管在处理文件过程中是否发生异常，都能保证 with 语句执行完毕后已经关闭了打开的文件句柄

     f.write(header.tobytes())

     #写入二进制数

     f.write(labeldata.tobytes())

 def write_imagedata(imagedata, outputfile):

   header = numpy.array([0x0803, len(imagedata), 28, 28], dtype='>i4')

   with open(outputfile, "wb") as f:

     f.write(header.tobytes())

     f.write(imagedata.tobytes())

 def main(argv):

   # Uncomment the line below if you want to seed the random

   # number generator in the same way I did to produce the

   # specific data files in this repo.

   # random.seed(int("notMNIST", 36))

   #当我们设置相同的seed，每次生成的随机数相同。如果不设置seed，则每次会生成不同的随机数

   labelsAndFiles = get_labels_and_files(argv[1], int(argv[2]))

   #随机排序

   random.shuffle(labelsAndFiles)

   imagedata, labeldata = make_arrays(labelsAndFiles)

   write_labeldata(labeldata, argv[3])

   write_imagedata(imagedata, argv[4])

 if __name__=='__main__':

 #Make a script both importable and executable

 #如果我们是直接执行某个.py文件的时候，该文件中那么”__name__ == '__main__'“是True

 #如果被别的模块import，__name__！='__main__'，这样main()就不会执行

   main(sys.argv)

curl -o notMNIST_small.tar.gz http://yaroslavvb.com/upload/notMNIST/notMNIST_small.tar.gz

curl -o notMNIST_large.tar.gz http://yaroslavvb.com/upload/notMNIST/notMNIST_large.tar.gz

tar xzf notMNIST_small.tar.gz

tar xzf notMNIST_large.tar.gz

python convert_to_mnist_format.py notMNIST_small  data/t10k-labels-idx1-ubyte data/t10k-images-idx3-ubyte

python convert_to_mnist_format.py notMNIST_large  data/train-labels-idx1-ubyte data/train-images-idx3-ubyte

gzip data/*ubyte