Autoencoder networks belong to the unsupervised learning category of methods, where labeled target values are not available. However, since autoencoders often use targets that are some form of input data, they can also be called self-supervised learning methods. In this chapter, we will learn how to apply autoencoder neural networks using Keras. We will cover three applications of autoencoders: dimension reduction, image denoising, and image correction. The examples in this chapter will use images of fashion items, images of numbers, and pictures containing people.
More specifically, in this chapter, we will cover the following topics:
- Types of autoencoders
- Dimension reduction autoencoders
- Denoising autoencoders
- Image correction autoencoders
