From the course: Computer Vision for Data Scientists

Unlock the full course today

Join today to access over 24,900 courses taught by industry experts.

Best practices for transfer learning

Best practices for transfer learning

From the course: Computer Vision for Data Scientists

Start my 1-month free trial Buy for my team

Best practices for transfer learning

“

- [Instructor] Let's now talk about some best practices for transfer learning. So when do you use transfer learning? Transfer learning is most effective when your dataset is smaller or less diverse than the dataset the pre-trained model was trained on, or when the features in your dataset are similar to those in the dataset used for pre-training your model. When your dataset is large and differs significantly from the pre-trained model's dataset, training a model from scratch may be more effective. How do you choose the right approach for transfer learning? While choosing between feature extraction and fine-tuning is based on the size of your dataset and the similarity between your target problem and the pre-trained model's original task, feature extraction is more appropriate for smaller datasets because it reduces overfitting and improves generalization by using the pre-trained model's features as input. Now, I want to talk…

Contents