MNIST数据可视化

代码

import numpy as np
import struct
from scipy.misc import imsave
from PIL import Image

'''
mnist数据格式
'''

def loadImageSet(filename):
    # 读取二进制文件
    binfile = open(filename, 'rb')
    buffers = binfile.read()

    # 取前4个整数，返回一个元组
    head = struct.unpack_from('>IIII', buffers, 0)
    magicNum = head[0]
    imgNum = head[1]
    width = head[2]
    height = head[3]

    # 取data数据，返回一个元组
    offset = struct.calcsize('>IIII')  # 定位到data开始的位置

    bits = imgNum * width * height  # data一共有60000*28*28个像素值
    bitsString = '>' + str(bits) + 'B'  # fmt格式：'>47040000B'

    imgs = struct.unpack_from(bitsString, buffers, offset)

    binfile.close()

   # imgs = np.reshape(imgs, [imgNum, width * height])  # reshape为[60000,784]型数组
    imgs = np.array(imgs).astype(np.uint8).reshape(imgNum, 1, width, height)
    return imgs, head

if __name__=="__main__":
    file1 = './mnist/train-images.idx3-ubyte'
    imgs, data_head = loadImageSet(file1)
    #测试，输出第一张图，没有用for导出全部图像
    img = Image.fromarray(imgs[1, 0, 0:28, 0:28])
    img.save("mnist.jpg")

mnist.jpg