Quick Start¶

Run tree crown segmentation from a trained samstars model bundle and one small input area.

1) Install the library¶

python -m pip install samstars

2) Prepare a model bundle¶

samstars does not bundle pretrained model weights. Train a model bundle for your own data first, or prepare one separately.

See Training for how to produce a bundle directory with samstars.train(...).

3) Segment a small input area¶

from pathlib import Path

from samstars import segment

segment(
    img=Path("/abs/path/image.tif"),
    chm=Path("/abs/path/chm.tif"),
    den=Path("/abs/path/first_return_density.tif"),
    model=Path("/abs/path/your_model_bundle"),
    out=Path("/abs/path/run_dir/segmentation_run"),
    storage="temp",
    device="cpu",
)

CPU is the supported baseline. See Installation before selecting "cuda" or "mps".

Use storage="persistent" when you want to keep proposals, prep artifacts, and other run state under out. Use storage="temp" when you only want final outputs left in out.

Outputs¶

crown polygons: /abs/path/run_dir/segmentation_run/crowns/crowns.gpkg

Large rasters¶

See Large Raster Runs.