Understanding transfer learning

The rest of this chapter will involve transfer learning techniques; therefore, we will spend this section explaining how transfer learning works within our architecture.

Getting ready

There are no dependencies required for this section.

How to do it...

This section walks through the steps for how transfer learning works:

1. Identify a pre-trained model that will be used as the training methodology that will be transferred to our chosen task. In our case, the task will be in identifying images of Messi and Ronaldo.
2. There are several available pre-trained models that can be used. The most popular ones are the following:
   1. Xception
   2. InceptionV3
   3. ResNet50
   4. VGG16
   5. VGG19
3. The features from the pre-trained convolutional neural network are extracted and saved for a certain set of images over several layers of filtering and pooling.
4. The final layer for the pre-trained convolutional neural network is substituted with the specific features that we are looking to classify based on our dataset...