MNIST data set label image reading

Reprinted: https://blog.csdn.net/opipa/article/details/51882875

Label noise annotation is used, FASHION MNIST data set

label file reading method:

import struct
from array import array
with open("train-labels-idx1-ubyte", "rb") as f:
    magic, size = struct.unpack(">II", f.read(8))
    labels = array("B", f.read())
    print magic, size, labels

Note: The fourth line cannot be deleted, the first eight are not labels

This is how to read image files (under ipython):

import struct
from array import array

with open("t10k-images-idx3-ubyte", "rb") as f:
    magic, size, rows, cols = struct.unpack(">IIII", f.read(16))
    print magic, size, rows, cols
    image_data = array("B", f.read())
    images = []
    for i in range(size):
        images.append([0] * rows * cols)

    for i in range(size):
        images[i][:] = image_data[i * rows * cols:(i + 1) * rows * cols]

Display the first 72 pictures:

import numpy as np
from PIL import Image
import matplotlib.pyplot as plt
%matplotlib inline

for i,img in enumerate(images):
    if i < 72:
        plt.subplot(9,8,i+1)
        img = np.array(img)
        img = img.reshape(rows,cols)
        img = Image.fromarray(img)
        plt.imshow(img, cmap='gray')
        plt.axis("off")
    else:
        break