How to use Transfer Learning?
Explore the fundamentals of transfer learning to build a Pokemon image classifier. Understand how to leverage pre-trained convolutional neural networks by using them as fixed feature extractors or by fine-tuning their layers for improved accuracy. This lesson guides you through practical strategies to adapt existing deep learning models to new image datasets efficiently.
We'll cover the following...
We'll cover the following...
How to use transfer learning?
Before moving to how we can leverage the transfer learning paradigm, we must know that almost all of the networks (either pre-trained or your own) have a common structure.
This is illustrated below:
Let’s understand the architecture shown above.
- The convolutional base is just a feature extractor that learns and extracts the features of any image. It contains the convolution and pooling layers.
- The Multi-layer Perceptron is the dense layer