Quick start

Three lines, any method.

Unsupervised

from raster2poly import RasterClassifier

clf = RasterClassifier("satellite.tif")
gdf = clf.unsupervised(n_clusters=6, algorithm="mini_batch_kmeans")
clf.save(gdf, "clusters.gpkg")

Supervised (ROI shapefile)

# If your labels are strings, encode them first:
out_path, mapping = clf.encode_roi("training.shp", label_col="LandCover")
#  1: Bare soil
#  2: Forest
#  3: Water
#  ...

gdf = clf.supervised(roi_path=out_path, class_col="class_id")
clf.save(gdf, "supervised.geojson")

Rule-based (DN ranges)

# Check band value ranges first
clf.band_stats()

rules = {
    1: [(4, 0.15, 1.0), (5, 0.0, 0.10)],  # high band 4, low band 5
    2: [(5, 0.25, 1.0)],                     # high band 5
}
gdf = clf.from_dn_ranges(rules)
clf.save(gdf, "dn_rules.gpkg")