Upload folder using huggingface_hub

per_country/README.md

@@ -0,0 +1,22 @@
+# per_country/
+
+One folder per European country (46 total). Each folder contains:
+
+- `<country>.parquet` — every polygon from that country that passed
+  the filter chain (see root README).
+- `README.md` — short country note: polygon count, extract status,
+  PBF date, and any caveats (e.g. regional sub-PBF processing for
+  France/Germany).
+
+The per-country split mirrors Geofabrik's regional extracts at
+`https://download.geofabrik.de/europe/`. Use this folder when you
+want a single country without paying for the full 9 GB combined file.
+
+Total polygons across all 46 countries: **7,112,375**.
+
+For the all-in-one file, see [`../combined/`](../combined/). For a
+small representative sample (~4k polygons, one per ~1.7k), see
+[`../sample/`](../sample/). For a thumbnail of the geographic
+distribution, see [`../preview/`](../preview/).
+
+[Back to the dataset root](../README.md)

per_country/albania/README.md

@@ -0,0 +1,36 @@
+# albania
+
+14,738 polygons from `albania-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/albania.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `albania` |
+| polygons | **14,738** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `albania-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/albania.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`albania.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/albania/albania.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Albania's OSM coverage has grown sharply since 2017; Tirana and the coastal strip are well-mapped. Source: Geofabrik Europe/Albania extract.

per_country/albania/albania.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebf9e6ea6bbccbf133e52fc133db4c7f41ff38b4ee7369a68dee90eb4680a7f2
+size 16596954

per_country/andorra/README.md

@@ -0,0 +1,36 @@
+# andorra
+
+776 polygons from `andorra-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/andorra.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `andorra` |
+| polygons | **776** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `andorra-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/andorra.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`andorra.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/andorra/andorra.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Tiny principality in the Pyrenees. The whole country fits in a single tile, so even the small extract yields good coverage of hiking trails, landuse, and buildings.

per_country/andorra/andorra.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c545e3e6657434a5e8fc958c443b6793744006570eb9a51710e2f9520059e7b0
+size 1997868

per_country/austria/README.md

@@ -0,0 +1,36 @@
+# austria
+
+133,711 polygons from `austria-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/austria.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `austria` |
+| polygons | **133,711** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-27 |
+| source | `austria-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/austria.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`austria.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/austria/austria.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Strong community mapping across all nine Bundesländer. Excellent coverage of landuse (agriculture, forest) and alpine hiking infrastructure.

per_country/austria/austria.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f1f91c83bc7f1a4f47c7006504ab2a43721833f8af9fced592717784da92450
+size 240044688

per_country/azores/README.md

@@ -0,0 +1,36 @@
+# azores
+
+2,640 polygons from `azores-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/azores.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `azores` |
+| polygons | **2,640** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `azores-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/azores.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`azores.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/azores/azores.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Portuguese archipelago in the Atlantic. Polygons cover the nine inhabited islands; remote islets are mostly absent from OSM.

per_country/azores/azores.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:776423c4cde46d31a6c2e10f2be03a8fe5719dc7feb66623132e9fb1071f35a5
+size 7576363

per_country/belarus/README.md

@@ -0,0 +1,36 @@
+# belarus
+
+223,750 polygons from `belarus-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/belarus.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `belarus` |
+| polygons | **223,750** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `belarus-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/belarus.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`belarus.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/belarus/belarus.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Mapping is active but uneven. Minsk and regional capitals have dense coverage; rural landuse is patchier.

per_country/belarus/belarus.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9dd35ef9152c97feb0148b733f234dba2a10ba6be0c013042625272b4824a05
+size 169284163

per_country/belgium/README.md

@@ -0,0 +1,36 @@
+# belgium
+
+125,108 polygons from `belgium-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/belgium.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `belgium` |
+| polygons | **125,108** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `belgium-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/belgium.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`belgium.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/belgium/belgium.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Dense, high-quality mapping across Flanders, Wallonia, and Brussels. Excellent for benchmarking against official cadastral data.

per_country/belgium/belgium.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ad22e96959b8092fc4f4f2c27c9e9e4a297ff6611e50b55868d3d8be3333ad4
+size 93474602

per_country/bosnia-herzegovina/README.md

@@ -0,0 +1,36 @@
+# bosnia-herzegovina
+
+49,715 polygons from `bosnia-herzegovina-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/bosnia-herzegovina.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `bosnia-herzegovina` |
+| polygons | **49,715** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `bosnia-herzegovina-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/bosnia-herzegovina.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`bosnia-herzegovina.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/bosnia-herzegovina/bosnia-herzegovina.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Coverage is solid in urban areas and along the main road network; mountain terrain (Dinaric Alps) is sparser.

per_country/bosnia-herzegovina/bosnia-herzegovina.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa9ac4362e51cb6cc8deae77d2ee5816cca91dc4e3023e31c0c3d6cc54c294ae
+size 109790500

per_country/bulgaria/README.md

@@ -0,0 +1,36 @@
+# bulgaria
+
+74,567 polygons from `bulgaria-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/bulgaria.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `bulgaria` |
+| polygons | **74,567** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `bulgaria-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/bulgaria.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`bulgaria.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/bulgaria/bulgaria.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Sofia, Plovdiv, and the Black Sea coast have dense landuse and building footprints; mountain areas (Rila, Pirin, Rhodopes) are better-mapped for hiking.

per_country/bulgaria/bulgaria.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0558a941411f11a1915c2d6749708424cdb82ac0374d5e8b302cf1a9f8324cf7
+size 88871644

per_country/croatia/README.md

@@ -0,0 +1,36 @@
+# croatia
+
+47,140 polygons from `croatia-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/croatia.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `croatia` |
+| polygons | **47,140** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `croatia-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/croatia.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`croatia.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/croatia/croatia.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Coast and islands are well-mapped for tourism; the interior (Slavonia) has solid agricultural landuse.

per_country/croatia/croatia.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:330da57275744813e91f4b6f71de16d8f20144e78771d8c32966fe1ab8e3c6fe
+size 88398398

per_country/cyprus/README.md

@@ -0,0 +1,36 @@
+# cyprus
+
+4,846 polygons from `cyprus-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/cyprus.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `cyprus` |
+| polygons | **4,846** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `cyprus-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/cyprus.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`cyprus.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/cyprus/cyprus.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+The divided island. OSM coverage is strong in the Republic of Cyprus; the north is mapped but partially via imports.

per_country/cyprus/cyprus.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93667f8594be4421b240889a8ea07916344710775ef54010bd8df2cb07f398b2
+size 6564199

per_country/czech-republic/README.md

@@ -0,0 +1,36 @@
+# czech-republic
+
+271,062 polygons from `czech-republic-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/czech-republic.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `czech-republic` |
+| polygons | **271,062** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-27 |
+| source | `czech-republic-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/czech-republic.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`czech-republic.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/czech-republic/czech-republic.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+One of the best-mapped countries in central Europe. Excellent address, landuse, and building data.

per_country/czech-republic/czech-republic.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c77a3df100e23e66bba7263e8a682db24fea705ae08d35d37a593b396ee1220d
+size 494416392

per_country/denmark/README.md

@@ -0,0 +1,36 @@
+# denmark
+
+175,795 polygons from `denmark-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/denmark.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `denmark` |
+| polygons | **175,795** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-27 |
+| source | `denmark-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/denmark.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`denmark.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/denmark/denmark.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Among the world's best-mapped countries. Comprehensive address data, full landuse, and very fresh updates.

per_country/denmark/denmark.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22904daa2a528fcc67a8d6ee0b8fa92f7ae27357453fc18f9c232d47b8e7d674
+size 91665060

per_country/estonia/README.md

@@ -0,0 +1,36 @@
+# estonia
+
+47,160 polygons from `estonia-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/estonia.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `estonia` |
+| polygons | **47,160** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `estonia-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/estonia.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`estonia.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/estonia/estonia.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Strong nationwide mapping, including detailed landuse and forestry. Tallinn has excellent building footprints.

per_country/estonia/estonia.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7b27be4f024295f8daf2738fe7732e8e51d0f5a5ab9399dc2ee2e01de74d66d
+size 43089925

per_country/faroe-islands/README.md

@@ -0,0 +1,36 @@
+# faroe-islands
+
+1,278 polygons from `faroe-islands-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/faroe-islands.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `faroe-islands` |
+| polygons | **1,278** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `faroe-islands-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/faroe-islands.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`faroe-islands.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/faroe-islands/faroe-islands.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Small, well-mapped archipelago. The 18 main islands are all covered.

per_country/faroe-islands/faroe-islands.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17ef717e61552125b8c9172708ca6efc8516ce46d5677a188eed78934ddcbada
+size 2637220

per_country/finland/README.md

@@ -0,0 +1,36 @@
+# finland
+
+427,870 polygons from `finland-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/finland.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `finland` |
+| polygons | **427,870** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-27 |
+| source | `finland-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/finland.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`finland.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/finland/finland.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Excellent nationwide mapping. Strong coverage of forests, lakes, and the very long coastline.

per_country/finland/finland.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:525bc601090ad4b88bcf250d3ea135a9ab12b7361d5be654fd117773c7453f37
+size 511922824

per_country/france/README.md

@@ -0,0 +1,36 @@
+# france
+
+492,538 polygons from `france-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/france.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `france` |
+| polygons | **492,538** |
+| extract_status | **clean** |
+| pbf_date | unknown |
+| source | `france-latest.osm.pbf` *processed via 26 Geofabrik regional sub-PBFs* |
+| Geofabrik extract | <https://download.geofabrik.de/europe/france.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`france.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/france/france.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Processed via 26 Geofabrik regional sub-PBFs (régions + overseas). The country PBF was removed in favor of the regional breakdown because the parent extract was too large to process in one pass.

per_country/france/france.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d727cde54c4a28e9f2b2ae8a5800d591e9ea09dfe88e2761a9369f0d056aa3e
+size 607844532

per_country/germany/README.md

@@ -0,0 +1,36 @@
+# germany
+
+1,131,888 polygons from `germany-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/germany.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `germany` |
+| polygons | **1,131,888** |
+| extract_status | **clean** |
+| pbf_date | unknown |
+| source | `germany-latest.osm.pbf` *processed via 16 Geofabrik regional sub-PBFs* |
+| Geofabrik extract | <https://download.geofabrik.de/europe/germany.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`germany.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/germany/germany.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Processed via 16 Geofabrik regional sub-PBFs (Bundesländer). Highest polygon count of any country in the dataset, reflecting the very active German OSM community.

per_country/germany/germany.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98ff6d412435f5d40e3483c39a658a2e4bb2572fa98ee059e424c812f38a458b
+size 852264667

per_country/greece/README.md

@@ -0,0 +1,36 @@
+# greece
+
+45,142 polygons from `greece-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/greece.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `greece` |
+| polygons | **45,142** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `greece-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/greece.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`greece.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/greece/greece.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Athens and Thessaloniki have excellent urban coverage. The islands and mainland mountain areas are well-mapped for hiking.

per_country/greece/greece.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0ef20f28240067a845ed89fec288275655c1709a4758f6b4d5df363cd5cf112
+size 71653740

per_country/guernsey-jersey/README.md

@@ -0,0 +1,36 @@
+# guernsey-jersey
+
+670 polygons from `guernsey-jersey-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/guernsey-jersey.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `guernsey-jersey` |
+| polygons | **670** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `guernsey-jersey-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/guernsey-jersey.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`guernsey-jersey.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/guernsey-jersey/guernsey-jersey.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Channel Islands. Two crown dependencies mapped as a single extract.

per_country/guernsey-jersey/guernsey-jersey.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4192dd92577edf44fbfcb850303cc1fc1abd47aac90fb8d97ad0962cd5878894
+size 532966

per_country/hungary/README.md

@@ -0,0 +1,36 @@
+# hungary
+
+77,569 polygons from `hungary-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/hungary.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `hungary` |
+| polygons | **77,569** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `hungary-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/hungary.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`hungary.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/hungary/hungary.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Budapest has dense urban mapping; the Great Plain (Alföld) has good agricultural landuse.

per_country/hungary/hungary.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:753a07f6928211d1d5afb9da5da199496356f5ec0898c391b05ccc96fc98555d
+size 78492332

per_country/iceland/README.md

@@ -0,0 +1,36 @@
+# iceland
+
+47,896 polygons from `iceland-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/iceland.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `iceland` |
+| polygons | **47,896** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `iceland-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/iceland.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`iceland.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/iceland/iceland.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Sparse but comprehensive: roads, landuse, and the small settled areas are all well-mapped. The highlands are intentionally not in OSM (no trails).

per_country/iceland/iceland.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5c02f14ee41c115e94ed8cc2f8a6db7789cba681201c59c1d1f097c91efbcf0
+size 62181839

per_country/isle-of-man/README.md

@@ -0,0 +1,36 @@
+# isle-of-man
+
+2,648 polygons from `isle-of-man-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/isle-of-man.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `isle-of-man` |
+| polygons | **2,648** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `isle-of-man-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/isle-of-man.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`isle-of-man.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/isle-of-man/isle-of-man.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+British Crown dependency. Small island, good coverage.

per_country/isle-of-man/isle-of-man.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aef05552ef9cb2bb13fdb905a7bd40c922ee2667f6877d76e96ec89b2a93a872
+size 1176072

per_country/italy/README.md

@@ -0,0 +1,36 @@
+# italy
+
+276,991 polygons from `italy-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/italy.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `italy` |
+| polygons | **276,991** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-27 |
+| source | `italy-latest.osm.pbf` *processed via 5 Geofabrik regional sub-PBFs* |
+| Geofabrik extract | <https://download.geofabrik.de/europe/italy.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`italy.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/italy/italy.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Processed via 5 Geofabrik regional sub-PBFs (centro, isole, nord-est, nord-ovest, sud). Strong urban mapping across all regions.

per_country/italy/italy.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cb898597cb052fd11badb710d7d897045af4add6f42bb41b9c81a06a50873dc
+size 587718752

per_country/kosovo/README.md

@@ -0,0 +1,36 @@
+# kosovo
+
+5,377 polygons from `kosovo-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/kosovo.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `kosovo` |
+| polygons | **5,377** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `kosovo-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/kosovo.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`kosovo.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/kosovo/kosovo.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Recognized by many OSM contributors as a separate territory. Polygons reflect the boundary used by Geofabrik.

per_country/kosovo/kosovo.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25c885fb0040ad029a397e1aec6d33573a5177c0997dc6613f8d123e698dfe3a
+size 6681122

per_country/latvia/README.md

@@ -0,0 +1,36 @@
+# latvia
+
+47,133 polygons from `latvia-latest.osm.pbf` on Geofabrik
+([source](https://download.geofabrik.de/europe/latvia.html)), filtered by the
+[22,075-tag whitelist](https://github.com/NoeFlandre/osm-stats) and
+classified by continent + size bin.
+
+| field | value |
+|-------|-------|
+| country | `latvia` |
+| polygons | **47,133** |
+| extract_status | **clean** |
+| pbf_date | 2026-06-26 |
+| source | `latvia-latest.osm.pbf` |
+| Geofabrik extract | <https://download.geofabrik.de/europe/latvia.html> |
+
+## Geometry
+
+The parquet file in this folder
+(`latvia.parquet`) has the same schema as the combined
+`combined/all_europe.parquet`. Each row carries the polygon
+**geometry as WKT** (default), plus centroid + area + the
+whitelist-matched tag.
+
+Load with:
+
+```python
+import pyarrow.parquet as pq
+table = pq.read_table("per_country/latvia/latvia.parquet")
+df = table.to_pandas()
+# df["geometry_wkt"] is a column of WKT strings; parse with shapely.wkt.loads
+```
+
+## Notes
+
+Solid nationwide coverage; Riga has detailed urban data.