Transfer Learning
Explore transfer learning techniques with PyTorch to reuse pretrained model weights for new image classification problems. Learn how to customize model layers, freeze parameters, and adapt architectures for different class sets to efficiently improve training outcomes.
We'll cover the following...
Structure of transfer learning
Transfer learning is the application of knowledge gained from completing one task to help solve a different but related problem. We discussed the different types of weight initialization. How about we use already trained and prepared weights to initialize our model? That’s the exact logic of transfer learning.
Imagine that we already have a model trained to predict cats and dogs, now we need another model to classify birds and horses. Instead of training a new model from scratch, we can use this previous model already learned how to extract significant features like legs, eyes, and many more cat and dog classes.
The interesting part is that ...