Summary

This chapter focused on CNNs, which consist of a kind of neural network architecture that performs outstandingly well on computer vision problems. We started by explaining the main reasons why CNNs are widely used for dealing with image datasets, as well as providing an introduction to the different tasks that can be solved through their use.

This chapter explained the different building blocks of a network's architecture by explaining the nature of convolutional layers, pooling layers, and, finally, FC layers. In each section, an explanation of the purpose of each layer was included, as well as code snippets that can be used to effectively code the architecture in PyTorch.

This led to the introduction of an image classification problem focused on classifying images of vehicles and animals. The purpose of this problem was to put the different building blocks of CNNs into practice to solve an image classification data problem.

Next, data augmentation was introduced...