geofeatures 0.1.0

Zonal statistics and NDVI feature extraction for remote sensing ML pipelines

GeoFeatures

Lightweight Python library for extracting zonal statistics and vegetation indices (NDVI) from satellite imagery for remote sensing ML pipelines.

Why

Computing per-polygon raster statistics (e.g. mean NDVI per farm parcel or administrative region) is a repetitive step in almost every remote sensing / geospatial ML workflow. GeoFeatures wraps this into a clean, tested API.

Install

pip install geofeatures

Quick example

from geofeatures.core import ndvi_zonal_stats
import geopandas as gpd

polygons = gpd.read_file("parcels.geojson")

result = ndvi_zonal_stats(
    nir_band=nir,
    red_band=red,
    vector_gdf=polygons,
    raster_template_path="ndvi.tif"
)

print(result[["name", "mean", "std"]])

Validated on real data

Tested end-to-end on Sentinel-2 L2A imagery (via Microsoft Planetary Computer) over Ogun State, Nigeria — mean NDVI of 0.29 across ~4M pixels, consistent with mixed agricultural/urban land cover.

API

• compute_ndvi(nir_band, red_band) — compute NDVI from raw bands
• extract_zonal_features(raster_path, vector_gdf, stats) — zonal statistics per polygon
• ndvi_zonal_stats(nir_band, red_band, vector_gdf, raster_template_path) — convenience wrapper combining both

License

MIT