osm-wikidata-worldcover

A supervised text to land-cover dataset. Each example pairs a Wikipedia or Wikivoyage article with the ESA WorldCover class that covers at least 80% of the OpenStreetMap polygon the article describes.

1,014,354 examples, 388,372 distinct polygons, 1,007,904 distinct documents.

from datasets import load_dataset

ds = load_dataset("NoeFlandre/osm-wikidata-worldcover")
print(ds["train"][0]["text"][:200], ds["train"][0]["worldcover_label"])

Geographic coverage map

The map shows the centroids of all 388,372 distinct polygons represented in this release and carrying an ESA WorldCover label. Each point is colored by its WorldCover class. Points are centroids, not polygon boundaries; use lat, lon, and centroid_wkt for the tabular locations.

Example labelled polygons

polygon name	ESA WorldCover class
"a budi"	Tree cover
2121	Shrubland
"Omonqo'ton" Davlat Buyurtma Qo'riqxonasi	Grassland
's Heeren Loo	Cropland
"Akkon Heidenau"	Built-up
"Jayron" ekomarkazi	Bare / sparse vegetation
"Watzmanngletscher"	Snow and ice
'Muela Reservoir	Permanent water bodies
Abborrsjön	Herbaceous wetland
Abavi Island	Mangroves
Abyss Pool	Moss and lichen

One deterministic named polygon is shown for each ESA WorldCover class represented in this release.

What the label means

WorldCover classifies the ground in 10 m pixels. The label therefore describes the land cover of the area containing a feature, not the feature itself: a church is Built-up because its surroundings are, not because the building was classified. Treat this as place-context classification.

Each row carries polygon_area_m2 and observed_fraction so you can restrict to polygons large enough for the dominance test to have been a real filter (polygon_area_m2 >= 2500 is 25+ pixels).

Splits

split	examples	polygons	documents
train	806,879	308,646	801,418
validation	101,661	40,134	101,174
test	105,814	39,592	105,312

Classes

code	class	examples	share
10	Tree cover	171,611	16.9%
20	Shrubland	1,523	0.2%
30	Grassland	83,039	8.2%
40	Cropland	6,910	0.7%
50	Built-up	625,196	61.6%
60	Bare / sparse vegetation	22,494	2.2%
70	Snow and ice	3,276	0.3%
80	Permanent water bodies	98,195	9.7%
90	Herbaceous wetland	1,106	0.1%
95	Mangroves	624	0.1%
100	Moss and lichen	380	0.0%

Languages (top 15)

language	examples	share
en	123,698	12.2%
de	92,571	9.1%
fr	64,721	6.4%
ceb	46,313	4.6%
sv	39,842	3.9%
it	38,646	3.8%
es	38,202	3.8%
arz	33,679	3.3%
pl	31,653	3.1%
ja	29,990	3.0%
ru	28,183	2.8%
nl	28,081	2.8%
zh	21,701	2.1%
uk	17,920	1.8%
pt	16,854	1.7%

Geographic coverage

• H3 cells (resolution 5): 54,752
• Source regions: 379
• Bounding box: -179.74357215692368, -56.53793355247686 to 179.95081025067793, 82.71920186674947

Guarantees

Every build is checked against these and fails if any breaks:

• No polygon leakage — no polygon appears in more than one split.
• No document leakage — no document appears in more than one split.
• Geographic blocking — splits are assigned per H3 cell, so nearby places cannot straddle train, validation and test.
• Valid labels — every label is one of the 11 real WorldCover classes; no-data is never a label.
• Dominance — every row's class covers at least 80% of its polygon.
• No exact duplicates — no two rows share both text and label.
• Deterministic — rebuilding the same inputs yields byte-identical files.

Columns

column	meaning
text	Full article text
worldcover_code / worldcover_label	The target class
dominant_fraction	Share of the polygon covered by that class
observed_fraction	Share of the polygon observed at all
polygon_id, osm_type, osm_id	The OpenStreetMap object
lat, lon, centroid_wkt	Polygon centroid
polygon_area_m2	Polygon area
language, project, title, url	The article
split, h3_cell	Split and the cell it was decided on

Provenance

• Source: NoeFlandre/osm-polygon-wikidata-and-wikipedia at revision f48c5aaaec6aecd63ecf5c195565cc2787597f1e
• Land cover: ESA WorldCover 2021 v200 (10 m)
• Dominance threshold: 0.8
• Minimum article length: 10 words
• Maximum polygon area: 10000000000.0 m2
• Split seed: 20180101, H3 resolution 5

Code: github.com/NoeFlandre/osm-wikidata-worldcover

Rows removed after labelling

reason	rows
documents_split_across_splits	638
duplicate_examples	103,496
duplicate_objects_across_regions	65,357

Licence

Article text is CC BY-SA 4.0 (Wikipedia/Wikivoyage). OpenStreetMap geometry is ODbL. ESA WorldCover is CC BY 4.0.