Transfer learning is amazing. In fact, in a book full of amazing things, it might be the most amazing thing I have to tell you about. If not, it's at least perhaps the most useful and pragmatic deep learning technique I can teach you. Transfer learning helps you solve deep learning problems, especially computer vision problems, with very little data and very little computational power relative the problem's scope. In this chapter, we're going to talk about what transfer learning is, when you should use it, and finally how to do transfer learning in Keras.

We will cover the following topics in this chapter:

• Overview of transfer learning
• When transfer learning should be used
• The impact of source/target volume and similarity
• Transfer learning in Keras