Models learning path

This page is a short route through the model workflow. The linked pages are the canonical instructions.

1. Choose a model type.
2. Prepare and finalize a compatible dataset version.
3. Create and train the model.
4. Interpret its training and validation metrics.
5. Adjust the model or reuse trained weights in a new draft version.
6. After selecting and locking a version, generate and interpret one test run on untouched Test data.
7. Run inference on eligible imagery after reviewing its processing provenance and wavelength coverage.

For an end-to-end example, follow the plastics walkthrough.