MNIST 可视化

MNIST数据可视化，我参考了http://blog.csdn.net/u014046170/article/details/47445919#python的代码，将它的两个函数的功能写到了一块。记录在这，更改后的代码记录在这作为备忘。

from PIL import Image
import struct,os

def _download(data_dir):
    if not os.path.isdir(data_dir):
        os.system("mkdir " + data_dir)
    os.chdir(data_dir)
    if (not os.path.exists('train-images-idx3-ubyte')) or \
       (not os.path.exists('train-labels-idx1-ubyte')) or \
       (not os.path.exists('t10k-images-idx3-ubyte')) or \
       (not os.path.exists('t10k-labels-idx1-ubyte')):
        os.system("wget http://webdocs.cs.ualberta.ca/~bx3/data/mnist.zip")
        os.system("unzip -u mnist.zip; rm mnist.zip")
    os.chdir("..")


def vision(train_data,train_label,save_dir):
    #image
    image_file = open(train_data,'rb')
    image_index = 0
    image_buf = image_file.read()
    image_file.close()

    image_magic, image_images, image_rows, image_columns = struct.unpack_from('>IIII', image_buf, image_index)
    image_index = struct.calcsize('>IIII')

    #label
    label_file = open(train_label,'rb')
    label_index = 0
    label_buf = label_file.read()
    label_file.close()

    label_magic, label_items = struct.unpack_from('>II',label_buf,label_index)
    label_index = struct.calcsize('>II')

    #handle
    for i in xrange(image_images):
        image=Image.new('L',(image_columns,image_rows))
        label=int(struct.unpack_from('>B',label_buf,label_index)[0])
        label_index += struct.calcsize('>B')

        for x in xrange(image_rows):
            for y in xrange(image_columns):
                image.putpixel((y,x),int(struct.unpack_from('>B',image_buf,image_index)[0]))
                image_index += struct.calcsize('>B')

        print 'save ' + str(i) + '_' + str(label) + ' image'
        image.save(save_dir + str(label) + '/' + str(i) + '.png')





#python ~/mxnet/tools/make_list.py train/ train --recursive=True --exts=.png   

if __name__ == '__main__':
    _download('MNIST')
    print 'Please copy down code and press Enter:\n'+'mkdir -vp MNIST/{train,test}/{0,1,2,3,4,5,6,7,8,9} && exit 0\n'
    os.system(" bash")
    vision('MNIST/train-images-idx3-ubyte','MNIST/train-labels-idx1-ubyte','MNIST/train/')
    vision('MNIST/t10k-images-idx3-ubyte','MNIST/t10k-labels-idx1-ubyte','MNIST/test/')
    os.system('python ~/mxnet/tools/make_list.py MNIST/train/ MNIST/train --recursive=True --exts=.png')
    os.system('~/mxnet/bin/im2rec MNIST/train.lst MNIST/train/ MNIST/train.res')
    os.system('python ~/mxnet/tools/make_list.py MNIST/test/ MNIST/test --recursive=True --exts=.png')
        os.system('~/mxnet/bin/im2rec MNIST/test.lst MNIST/test/ MNIST/test.res')

用法向这样：

user@debian: python
>>> import test #上面的文件保存到test.py文件中
>>> test.vision('train-images-idx3-ubyte','train-labels-idx3-ubyte')  #这样就在当前目录下的test文件夹中生成了相关的数字文件