georeadr

Know if your geospatial data is ready for AI/ML before you find out the hard way.

pip install georeadr

Requires Python 3.9+ · Depends on geopandas

The problem

You have a shapefile or GeoPackage. You want to feed it into a model. But is the data actually ready?

• Does it have a CRS?
• Are there null geometries hiding in there?
• Are coordinates swapped or out of bounds?
• Are attributes complete enough to be useful?
• Are there duplicate features that will bias your model?

Finding out halfway through a pipeline is painful. georeadr checks first.

Quick start

import georeadr

report = georeadr.check("my_data.gpkg")
print(report)

Output:

============================================================
  georeadr AI/ML Readiness Report
  File: my_data.gpkg
============================================================
  Score: 74.2/100   Grade: C
  Dataset is mostly ready but has 2 warning(s) to address.
============================================================

[PASS] CRS defined
       CRS is defined: EPSG:27700

[PASS] Null geometries
       No null or empty geometries found.

[PASS] Geometry validity
       All geometries are valid.

[WARN] Attribute completeness
       Attribute completeness: 87.3%. 2 column(s) have missing values.

[PASS] Geometry type consistency
       All geometries are of type: Polygon

[WARN] Duplicate features
       3 duplicate geometries found (1.2%).

[PASS] Feature count
       2,847 features found.

[PASS] Bounds sanity
       Bounds look sensible (width=3.2145, height=2.8721).
============================================================

Works with any GIS format

georeadr.check("data.shp")
georeadr.check("data.geojson")
georeadr.check("data.gpkg")
georeadr.check("data.gpkg", layer="my_layer")

Or pass a GeoDataFrame directly:

import geopandas as gpd
import georeadr

gdf = gpd.read_file("data.shp")
report = georeadr.check(gdf)

Run specific checks only

report = georeadr.check("data.gpkg", checks=["crs", "nulls", "validity"])

Available checks: crs, nulls, validity, completeness, duplicates, geometry_type, count, bounds

Use the report programmatically

report = georeadr.check("data.gpkg")

report.score        # 74.2
report.grade        # "C"
report.passed       # True (score >= 70)
report.errors       # list of failed checks with severity="error"
report.warnings     # list of failed checks with severity="warning"

for check in report.checks:
    print(f"{check.name}: {'PASS' if check.passed else 'FAIL'} ({check.score:.0%})")

Checks explained

License

MIT