AI and ML for Coders in PyTorch

Chapter 2. Introduction to Computer Vision

Chapter 1 introduced the basics of how machine learning works. You saw how to get started with programming using neural networks to match data to labels, and from there, you saw how to infer the rules that can be used to distinguish items.

In this chapter, we’ll consider the next logical step, which is to apply these concepts to computer vision. In this process, a model learns how to recognize content in pictures so it can “see” what’s in them. You’ll work with a popular dataset of clothing items and build a model that can differentiate between them and thus “see” the difference between different types of clothing.

How Computer Vision Works

Computer vision is the ability of a computer to recognize items beyond just storing their pixels. For example, consider items of clothing that might look like those in Figure 2-1. They’re very complex, with lots of different varieties of the same item. Take a look at the two shoes—they’re very different, but they’re still shoes!

Figure 2-1. Clothing examples

There are a number of different recognizable clothing items here. You understand the difference between a shirt, a coat, and a dress, and you fundamentally know what each of these items are—but how would you explain all that to somebody who has never seen clothing? How about a shoe? There are two shoes in this image, but given the major differences ...