For this task, we will be using SVHN dataset, which is an abbreviation for Street View House Numbers by Stanford (http://ufldl.stanford.edu/housenumbers/). So, let's start the implementation by importing the required packages for this implementation:
# Lets start by loading the necessary libraries%matplotlib inlineimport pickle as pklimport timeimport matplotlib.pyplot as pltimport numpy as npfrom scipy.io import loadmatimport tensorflow as tfimport os
Next up, we are going to define a helper class to download the SVHN dataset (remember that you need to manually create the input_data_dir first):
from urllib.request import urlretrievefrom os.path import isfile, isdirfrom tqdm import tqdminput_data_dir = 'input/' ...