6 Transfer Learning

 

“Share your knowledge. It’s a way to achieve immortality”

-- Dalai Lama

Learning goals from this chapter

  • Understand transfer learning technique
  • Understand how transfer learning works
  • Learn the different transfer learning approaches
  • Learn how to choose the learning approach for your problem
  • Explore some of the most popular open-source image datasets that you can download and train your model for your own task
  • Build two end-to-end transfer learning projects using different transfer learning approaches

When you're building a computer vision application, you can build your convnets as we learned in chapter three and start the training from scratch. And that is an acceptable approach. Another much faster approach is to download a neural network that someone else has already built and trained on a large dataset in a certain domain and use this pretrained network as a starting point to train the network on your new task. This approach is called transfer learning.

6.1   What are the problems that transfer learning is solving?

6.2   What is transfer learning?

6.3   How transfer learning works

6.3.1   How do neural networks learn features?

6.3.2   What about the transferability of features extracted at later layers in the network?

6.4   Transfer learning approaches

6.4.1   Pretrained network as a classifier

6.4.2   Pretrained network as a feature extractor

6.4.3   Fine-tuning

6.5   Choose the appropriate level of transfer learning

6.5.1   Scenario #1: target dataset is small and similar to source dataset

6.5.2   Scenario #2: target dataset is large and similar to the source dataset

6.5.3   Scenario #3: target dataset is small and different from the source dataset

sitemap