Introduction

In the previous chapter, you learned how convolution neural networks (CNNs) analyze images and learn relevant patterns to classify their main subjects or identify objects within them. You also saw the different types of layers used for such models.

But rather than training a model from scratch, it would be more efficient if you could reuse existing models with pre-calculated weights. This is exactly what transfer learning and fine-tuning are about. You will learn how to apply these techniques to your own projects and datasets in this chapter.

You will also look at the ImageNet competition and the corresponding dataset that is used by deep learning researchers to benchmark their models against state-of-the-art algorithms. Finally, you will learn how to use TensorFlow Hub's resources to build your own model.