patatahooligan / Mnist-data-numpy-format

Mnist-data-numpy-format

This data is originally from the mnist database. I have formatted it in the following way for ease of use in my tutorial series:

mnist.npz contains 6 files:

'training_images', 'training_labels', 'test_images', 'test_labels', 'validation_images', 'validation_labels'

The image files are each stored as an array of 784x1 arrays.

The label files are each stored as an array of 10x1 arrays (one-hot encoding)

The data can be loaded like so (assumes mnist.npz is in same folder as your script):

import numpy as np

with np.load('mnist.npz') as data: 
	training_images = data['training_images']

Depending on your setup you may have to use the absolute path to the file, which can be done like so:

import numpy as np
import os

path = os.path.join(os.path.dirname(os.path.realpath(__file__)), 'mnist.npz')
with np.load(path) as data: 
	training_images = data['training_images']