This chapter showed us how to make a CNN for classifying images in the CIFAR-10 dataset. The classification accuracy was about 79% - 80% on the test set. The output of the convolutional layers was also plotted, but it was difficult to see how the neural network recognizes and classifies the input images. Better visualization techniques are needed.

Next up, we'll use one of the modern and exciting practice of deep learning, which is transfer learning. Transfer learning allows you to use data-greedy architectures of deep learning with small datasets.