Lesson 9: Urban Growth with Machine Learning¶
Planning question: How has the city expanded?
East Bengaluru has grown rapidly along its IT corridors. You will train a Random Forest classifier with ESA WorldCover labels, map land cover in 2019 and 2024, and measure the new built-up area.
What you will learn¶
- Turn satellite images into a table of pixels and features
- Train and test a Random Forest classifier
- Read precision, recall and the confusion matrix
- Classify two years and map the change
Key concepts¶
- Supervised classification: learning from labelled examples
- Training and test sets: checking the model on unseen data
- Change detection: comparing classified maps over time
What you will make¶
Land cover maps for 2019 and 2024, and a map of new built-up areas.
| Packages | scikit-learn, odc-stac, xarray |
| Data | Sentinel-2 L2A, ESA WorldCover 2021 |
| Time | 30 to 45 minutes |
Try it yourself¶
The end of the notebook includes these exercises:
- Study the airport area.
- Train without the indices.
- Use fewer trees.
Enjoying the course? Support it with a star on GitHub.