In this section, we'll compare our models in the font classification problem. First, we should remind ourselves what the data looks like. Then, we'll inspect the simple logistic dense neural network and convolutional neural network models. You've come a long way in modeling with TensorFlow.

Before we move on from deep learning, however, let's look back and see how models compare on the font classification problem. First, let's look at the data again, so we don't lose sight of the problem. In fact, let's look at one image that includes all the letters and digits from every font, just to see what shapes we have:

# One look at a letter/digit from each font
# Best to reshape as one large array, then plot
all_letters = np.zeros([5*36,62*36])
for font in range(5):
    for letter in range(62):
        all_letters[font*36:(font+1)*36,
                letter*36:(letter+1)*36] = 
                train[9*(font*62 + letter)]

This would be a lot of subplots for Matplotlib to handle. So, we'll make a new array, 5 images tall, 5 fonts times 36 pixels, and 62 images wide, 62 letters or digits times 36 pixels. After allocating a zero array, we can stack the training images into it. Fonts and letters act as indices and we set 36x36 values at a time in the big array. Note that we have 9 in the train array here, because we're only taking one type of jitter each letter.

Let's take a look with a quick call to pcolormesh:

plt.pcolormesh(all_letters,
        cmap=plt.cm.gray)

As you can see, we have the entire alphabet, upper and lowercase, as well as digits 0 to 9. Some of the fonts look like others, while font 0 is kind of in its own world, to human eyes anyway. Each font has interesting stylistic properties that we hope our model picks up on.