Application to image denoising

Training our previous model for image denoising is as simple as creating a noisy copy of the training images and passing it as input to our network instead:

x_noisy = x_train + np.random.normal(loc=.0, scale=.5, size=x_train.shape)
autoencoder.fit(x_noisy, x_train)

The first two notebooks dedicated to this chapter detail the training process, providing illustrations and additional tips (for instance, to visualize the images predicted during training).