Examples¶
These notebooks demonstrate workflows using your own imagery, lidar rasters, and annotations. Replace the paths in each notebook before running it. The source data and base checkpoint are supplied separately; see Data Preparation and Training.
Open a notebook with a Python environment containing samstars and Jupyter.
The code uses the installed package and can run outside a repository checkout.
| Notebook | Purpose |
|---|---|
| Train and Segment One Area | Train on a labeled area and segment another area. |
| Train and Segment a Tiled Area | Train a bundle, prepare larger rasters as tiles, and segment them. |
| Retrain Refinement for One Area | Reuse saved proposals to train and run a replacement refinement model. |
The first two notebooks show the essential settings first, with optional tuning at the end. For evaluation, use areas held out from training; the notebooks demonstrate execution of the workflow.
See Quick Start for a compact example and Advanced Workflows for intermediate files and partial reruns.