Data Science Programming All-in-One For Dummies

Chapter 3

Recognizing Images with CNNs

IN THIS CHAPTER

Performing basic image recognition

Working with convolutions

Looking for edges and shapes in images

Humans are graphically (visually) oriented, but computers aren’t. So, the task that humans most want performed by deep learning — recognizing items in images — is also one of the harder tasks to perform using a computer. The manner in which a computer deals with images is completely different from humans. When working with images, computers deal with the numeric values that make up individual pixels. The computer processes the numbers used to create an image much as it processes any other group of numbers. Consequently, this chapter deals with using a different kind of math to manipulate those pixel values so that a computer can output the desired result, despite having no concept whatsoever that it is even processing an image.

You have some methods of working with images that don’t involve heavy-duty usage of deep learning techniques, but the output from these methods is also simple. However, these techniques make a good starting point for the discussions later in the chapter, so you see them in the first part of this chapter. ...