Exploring the MNIST dataset

Let's show a short example of how to access the MNIST data and how to display a selected image.

We import the following libraries:

The numpy library because we have to do some image manipulation:

 >>import numpy as np

The pyplot function in matplotlib for drawing the images:

>>import matplotlib.pyplot as plt

Finally, the mnist_data library, which you can download from the code repository of this book. It is a Google script that allows us to download the MNIST database and to build the dataset:

 >>import mnist_data 

 

Then we load the dataset using the read_data_sets method:

 >>__input = mnist_data.read_data_sets("data")

The data is the name of the directory where the images will be uploaded.

The shape of the images and labels:

>>__input.train.images.shape 
(60000, 28, 28, 1) 

>>__input.train.labels.shape 
(60000, 10) 

>>__input.test.images.shape 
(10000, 28, 28, 1) 

>>__input.test.labels.shape 
(10000, 10) 

Using the Python...