module_01 / benchmark configuration

Configure benchmark

1.1Datasets

Reference dataset

drop or choose a file · csv / geojson / parquet / gpkg

Comparable dataset(s)

drop or choose one or more files

Required fields: poi_id, name, address (street line only) and point coordinates in EPSG:4326 — a geometry, or lat / lon columns in CSV. The reference also needs category. Examples: reference.csv · comparable.csv

1.2Study area

1.3POI categories

select all clear default

Derived from the reference dataset. facilities and nature are off by default: large area-like POIs that point-to-point matching does not handle well.

1.4City center

Origin for the distance-to-center analysis (2 km bins).

1.5Projected CRS

Local UTM zone for distance calculation

Auto picks the UTM zone of the city center, anywhere in the world (north or south). Any projected CRS in meters also works. Used for location accuracy and distance to center; Stage-1 matching features keep EPSG:3857, as in the pretrained model.

study area

city center

$ cat config.json[ – ]