The problem with traditional deep neural networks

Before we dive into CNNs, let’s look at the major problem that’s faced when using traditional deep neural networks.

Let’s reconsider the model we built on the Fashion-MNIST dataset in Chapter 3. We will fetch a random image and predict the class that corresponds to that image, as follows:

The following code can be found in the Issues_with_image_translation.ipynb file located in the Chapter04 folder on GitHub at https://bit.ly/mcvp-2e. Only the additional code corresponding to the issue of image translation will be discussed here for brevity.

1. Fetch a random image from the available training images:

   # Note that you should run the code in
   # Batch size of 32 section in Chapter 3
   # before running the following code
   import matplotlib.pyplot as plt
   %matplotlib inline
   # ix = np.random.randint(len(tr_images))
   ix = 24300
   plt.imshow(tr_images[ix], cmap='gray')
   plt.title(fmnist...