Implementing a convolutional autoencoder

In the previous chapter, we how to implement an autoencoder for the Street View House Numbers dataset. We got some decent results, but the output could definitely be improved. In the following recipe, we will show how a convolutional autoencoder produces better outputs.

How to do it...

1. Let's start with importing the libraries as follows:

import numpy as np
import scipy.io

from matplotlib import pyplot as plt
from keras.utils import np_utils
from keras.models import Sequential, Input, Model
from keras.layers.core import Dense, Dropout, Activation, Reshape, Flatten, Lambda
from keras.layers import Conv2D, MaxPooling2D, UpSampling2D
from keras.callbacks import EarlyStopping

2. Next, we load the dataset and extract the data we will use in this recipe:

mat = scipy.io.loadmat('Data/train_32x32.mat')
mat = mat['X']
b, h, d, n = mat.shape

3. Before feeding the data to our network, we pre-process the data:

#Convert all RGB-Images to greyscale
img_gray = np.zeros(shape...