In this chapter, we're going to explore the vast and awesome world of computer vision.

If you've ever wanted to construct a predictive machine learning model using image data, this chapter will serve as an easily-digestible and practical resource. We'll go step by step through building an image-classification model, cross-validating it, and then building it in a better way. At the end of this chapter, we'll have a darn good model and discuss some paths for future enhancement.

Of course, some background in the fundamentals of predictive modeling will help this to go smoothly. As you'll soon see, the process of converting images into usable features for our model might might feel new, but once our features are extracted, the model-building and cross-validation processes are exactly the same.

In this chapter, we're going to build a convolutional neural network to classify images of articles of clothing from the Zalando Research dataset—a dataset of 70,000 images, each depicting 1 of 10 possible articles of clothing such as T-shirt/top, a pair of pants, a sweater, a dress, a coat, a sandal, a shirt, a sneaker, a bag, or an ankle boot. But first, we'll explore some of the fundamentals together, starting with image-feature extraction and walking through how convolutional neural networks work.  

So, let's get started. Seriously!.

Here's what we'll cover in this chapter:

• Image-feature extraction
• Convolutional neural networks:
  • Network topology
  • Convolutional layers and filters
  • Max pooling layers
  • Flattening
  • Fully-connected layers and output
• Building a convolutional neural network to classify images in the Zalando Research dataset, using Keras