Dataset Viewer
Auto-converted to Parquet Duplicate
id
large_stringlengths
32
32
text
large_stringlengths
2
186k
landuse
bool
2 classes
language
large_stringclasses
95 values
source
large_stringclasses
3 values
polygon_ids
large listlengths
1
3.57k
region
large_stringclasses
386 values
continent
large_stringclasses
8 values
lat
float64
-83.84
83.7
lon
float64
-179.99
180
name
large_stringlengths
0
255
⌀
osm_tag
large_stringlengths
7
141
⌀
area_m2
float64
0
10,160B
url
large_stringlengths
5
255
license
large_stringclasses
3 values
2fa0fe970af15f860f44595f89dcc4c4
No Roof.
true
afr_Latn
osm_description
[ "way/1074082042" ]
afghanistan
asia
33.11756
69.606888
null
building=construction
198.445511
https://www.openstreetmap.org/way/1074082042
ODbL-1.0
f5d65b58af247a108956d815e1847ab4
Kies und Ol
true
afr_Latn
osm_description
[ "way/25951247" ]
alsace
europe
48.647818
7.859347
Hafen Honau
natural=water
44,556.358502
https://www.openstreetmap.org/way/25951247
ODbL-1.0
c457bee2216976b841951368c959e854
Destrpyed
false
afr_Latn
osm_description
[ "way/427654580" ]
american-oceania
oceania
-14.349652
-170.758146
null
building=yes
122.276501
https://www.openstreetmap.org/way/427654580
ODbL-1.0
f5785892e2b2c89b85924b455d5a0371
Bullring
true
afr_Latn
osm_description
[ "way/172956262" ]
andalucia
europe
37.926517
-5.759845
Plaza de toros de Cazalla
historic=monument
766.141992
https://www.openstreetmap.org/way/172956262
ODbL-1.0
99e5c6fd9a27ad827b77a1143c4199e5
Restaurante "Los Manueles" - Self Service
false
afr_Latn
osm_description
[ "way/205384508" ]
andalucia
europe
37.796079
-3.72511
Los Manueles
amenity=restaurant
5,143.532888
https://www.openstreetmap.org/way/205384508
ODbL-1.0
25e8199ad2d98f341a40c18d26ce5b5c
Parking Burger King Estepona.
true
afr_Latn
osm_description
[ "way/541625602" ]
andalucia
europe
36.408179
-5.1892
Parking Burger King Estepona.
amenity=parking
756.300805
https://www.openstreetmap.org/way/541625602
ODbL-1.0
b10b6143dd57b4d271eac1060a159fe5
Old Water Plant
true
afr_Latn
osm_description
[ "way/329705245" ]
antarctica
antarctica
-77.850313
166.682049
null
building=warehouse
262.022312
https://www.openstreetmap.org/way/329705245
ODbL-1.0
ee7dc6fc4857a8310a56d99d4c1efa31
Parking sauvage
true
afr_Latn
osm_description
[ "way/736545596" ]
aquitaine
europe
44.901102
-0.212461
null
amenity=parking
84.223256
https://www.openstreetmap.org/way/736545596
ODbL-1.0
02390ed1d40437e0a36fea389899e184
Parking visiteurs
true
afr_Latn
osm_description
[ "way/1117159755" ]
aquitaine
europe
43.488862
-1.418505
null
amenity=parking
311.934109
https://www.openstreetmap.org/way/1117159755
ODbL-1.0
c4af5a7f8f82f08475ff7d8b3245c026
vente d'œufs
false
afr_Latn
osm_description
[ "way/67984680" ]
aquitaine
europe
45.542976
0.797444
La Fontaine aux OEufs
shop=farm
74.935244
https://www.openstreetmap.org/way/67984680
ODbL-1.0
6827d01e5fb2c5164aa8af0276bd9f6e
Parking sauvage
true
afr_Latn
osm_description
[ "way/736545597" ]
aquitaine
europe
44.901349
-0.212675
null
amenity=parking
104.763033
https://www.openstreetmap.org/way/736545597
ODbL-1.0
e0389b289893b6fb609d30b2733ad364
parking de la boulangerie
true
afr_Latn
osm_description
[ "way/797236656" ]
aquitaine
europe
43.77849
-1.21754
null
amenity=parking
107.939108
https://www.openstreetmap.org/way/797236656
ODbL-1.0
1449866794b79fbc1fa551539613b977
Reservorio
true
afr_Latn
osm_description
[ "way/1081803187" ]
argentina
south-america
-32.765499
-60.775773
null
natural=water
2,819.957757
https://www.openstreetmap.org/way/1081803187
ODbL-1.0
b9e833954c4f1fd4b58967f9b0ee8a25
DI.PR.E.GE.P. N° 5724 DI.PR.E.GE.P. N° 2853
false
afr_Latn
osm_description
[ "way/614367900" ]
argentina
south-america
-34.670408
-58.712422
Instituto Comandante Luis Piedrabuena
amenity=school
2,052.797501
https://www.openstreetmap.org/way/614367900
ODbL-1.0
ed9d747440ae5b9b544d2e67aa1f43eb
Gimnasio
true
afr_Latn
osm_description
[ "way/1464958535" ]
argentina
south-america
-32.922318
-60.81989
null
building=construction
45.465332
https://www.openstreetmap.org/way/1464958535
ODbL-1.0
4ef6dbe8e86208cb887afb02b75f6a97
Dapsa Agro
true
afr_Latn
osm_description
[ "way/664709751" ]
argentina
south-america
-35.888673
-57.889002
null
landuse=industrial
9,838.084177
https://www.openstreetmap.org/way/664709751
ODbL-1.0
2ca100dce4def0cba5bb4db26bb63bf0
#OpenStreetMap16
false
afr_Latn
osm_description
[ "way/835294943" ]
argentina
south-america
-27.352788
-55.896555
null
building=house
151.577885
https://www.openstreetmap.org/way/835294943
ODbL-1.0
a46aa72c0232c6cfcdd69de99a7fb038
Gimnasio
true
afr_Latn
osm_description
[ "way/473661648" ]
argentina
south-america
-34.471428
-58.512285
Los Druidas
leisure=sports_centre
121.875817
https://www.openstreetmap.org/way/473661648
ODbL-1.0
1605704ff451fd6ff5e986e07112bcc2
FAA: DOR
false
afr_Latn
osm_description
[ "way/961426163" ]
argentina
south-america
-38.741369
-61.256434
Aeródromo de Coronel Dorrego
aeroway=aerodrome
1,245,200.698947
https://www.openstreetmap.org/way/961426163
ODbL-1.0
b5d7b66cf5fcc99ad834490314fecef1
gimnasio
true
afr_Latn
osm_description
[ "way/1124963580" ]
argentina
south-america
-39.899492
-71.133487
null
leisure=sports_hall
654.526345
https://www.openstreetmap.org/way/1124963580
ODbL-1.0
ed202224d1b0d68122e96e6d6d4afba8
Reservorio
true
afr_Latn
osm_description
[ "way/1044488178" ]
argentina
south-america
-33.011073
-60.755516
null
natural=wetland
6,216.225894
https://www.openstreetmap.org/way/1044488178
ODbL-1.0
3b8acf28764530fc89e97bb03c348f27
rooftop
true
afr_Latn
osm_description
[ "way/326644005" ]
australia
oceania
-42.886399
147.313682
null
building=yes
112.92156
https://www.openstreetmap.org/way/326644005
ODbL-1.0
e97db2e1c9c6a9ebf3ffca19b4aa6ea5
rooftop
true
afr_Latn
osm_description
[ "way/326644114" ]
australia
oceania
-42.887625
147.31985
null
building=yes
212.467448
https://www.openstreetmap.org/way/326644114
ODbL-1.0
dc214bd85d62f783f4356c28f8daa093
rooftop
true
afr_Latn
osm_description
[ "way/326644249" ]
australia
oceania
-42.890849
147.310806
null
building=yes
41.107228
https://www.openstreetmap.org/way/326644249
ODbL-1.0
f4fdf488c379a567fd47bb5a144a4a3a
rooftop
true
afr_Latn
osm_description
[ "way/326644323" ]
australia
oceania
-42.8877
147.311961
null
building=yes
23.218501
https://www.openstreetmap.org/way/326644323
ODbL-1.0
8a45c43392089d8e8bbf1f92a318b051
rooftop
true
afr_Latn
osm_description
[ "way/326643830" ]
australia
oceania
-42.890011
147.312327
null
building=yes
152.640362
https://www.openstreetmap.org/way/326643830
ODbL-1.0
9dfaeb2b26350f2b60b96fa95d39c469
rooftop
true
afr_Latn
osm_description
[ "way/326644521" ]
australia
oceania
-42.886336
147.316979
null
building=yes
39.314685
https://www.openstreetmap.org/way/326644521
ODbL-1.0
2946072e4d4cb5fb8c07a28d40194c20
rooftop
true
afr_Latn
osm_description
[ "way/326644408" ]
australia
oceania
-42.887308
147.321055
null
building=yes
14.003574
https://www.openstreetmap.org/way/326644408
ODbL-1.0
4e5e1956767b29a5d3730314412784fc
rooftop
true
afr_Latn
osm_description
[ "way/326643975" ]
australia
oceania
-42.888135
147.317892
null
building=yes
183.761697
https://www.openstreetmap.org/way/326643975
ODbL-1.0
2227976d7731e3213055b6080e1f61da
rooftop
true
afr_Latn
osm_description
[ "way/326644193" ]
australia
oceania
-42.889572
147.314584
null
building=yes
161.817543
https://www.openstreetmap.org/way/326644193
ODbL-1.0
ff52477caa6091d52672dde7f4f03369
rooftop
true
afr_Latn
osm_description
[ "way/326643780" ]
australia
oceania
-42.8869
147.318123
null
building=yes
27.892483
https://www.openstreetmap.org/way/326643780
ODbL-1.0
11db6559a8749c895707d204f09f9d1f
rooftop
true
afr_Latn
osm_description
[ "way/326645311" ]
australia
oceania
-42.891714
147.317459
null
building=yes
318.834604
https://www.openstreetmap.org/way/326645311
ODbL-1.0
fc17ed99ca94c4c4a07c33d47f697a2e
rooftop
true
afr_Latn
osm_description
[ "way/326644518" ]
australia
oceania
-42.890286
147.321318
null
building=yes
8.717684
https://www.openstreetmap.org/way/326644518
ODbL-1.0
28959239721e9c288abe35bacc14923f
rooftop
true
afr_Latn
osm_description
[ "way/326643958" ]
australia
oceania
-42.888763
147.311689
null
building=yes
117.193034
https://www.openstreetmap.org/way/326643958
ODbL-1.0
a6f109355ac32f71064f9eb9c607dabf
rooftop
true
afr_Latn
osm_description
[ "way/326644555" ]
australia
oceania
-42.888058
147.312138
null
building=yes
87.863504
https://www.openstreetmap.org/way/326644555
ODbL-1.0
02d2a6f43fbf94df0fbaa102cb135c73
rooftop
true
afr_Latn
osm_description
[ "way/326644334" ]
australia
oceania
-42.887871
147.315995
null
building=yes
91.872451
https://www.openstreetmap.org/way/326644334
ODbL-1.0
637263345936d7ed6bfedcca1a3d4215
rooftop
true
afr_Latn
osm_description
[ "way/326643927" ]
australia
oceania
-42.889409
147.321249
null
building=yes
106.172723
https://www.openstreetmap.org/way/326643927
ODbL-1.0
1ce05cba4fb6e68612055f1c3dbffe8f
rooftop
true
afr_Latn
osm_description
[ "way/326644170" ]
australia
oceania
-42.88696
147.315844
null
building=yes
212.437611
https://www.openstreetmap.org/way/326644170
ODbL-1.0
06337a430b9a270389eb7df90375b75d
rooftop
true
afr_Latn
osm_description
[ "way/326643946" ]
australia
oceania
-42.887541
147.310944
null
building=yes
97.954932
https://www.openstreetmap.org/way/326643946
ODbL-1.0
03d402037922ffecc0ccb1dba5bd9729
rooftop
true
afr_Latn
osm_description
[ "way/326644740" ]
australia
oceania
-42.886852
147.310639
null
building=yes
118.589153
https://www.openstreetmap.org/way/326644740
ODbL-1.0
769efd62077fdb4f4274501931ce5cf4
rooftop
true
afr_Latn
osm_description
[ "way/326644087" ]
australia
oceania
-42.888653
147.31652
null
building=yes
180.363003
https://www.openstreetmap.org/way/326644087
ODbL-1.0
7fd8055d6d43dabd781e596f3eb175f7
rooftop
true
afr_Latn
osm_description
[ "way/326643879" ]
australia
oceania
-42.888965
147.318962
null
building=yes
105.073093
https://www.openstreetmap.org/way/326643879
ODbL-1.0
5b0805a78264ca7093a218f2488be399
rooftop
true
afr_Latn
osm_description
[ "way/326644686" ]
australia
oceania
-42.890803
147.311049
null
building=yes
39.261701
https://www.openstreetmap.org/way/326644686
ODbL-1.0
0747899bbeaa1938286b489dc088cb68
rooftop
true
afr_Latn
osm_description
[ "way/326644695" ]
australia
oceania
-42.887119
147.321125
null
building=yes
66.085213
https://www.openstreetmap.org/way/326644695
ODbL-1.0
2994d3a7d2103f27bf4bfe1116114419
rooftop
true
afr_Latn
osm_description
[ "way/326644575" ]
australia
oceania
-42.887283
147.314125
null
building=yes
13.154531
https://www.openstreetmap.org/way/326644575
ODbL-1.0
28a002ca697ebc8b4487421d25807d50
rooftop
true
afr_Latn
osm_description
[ "way/326643801" ]
australia
oceania
-42.88961
147.321039
null
building=yes
74.41595
https://www.openstreetmap.org/way/326643801
ODbL-1.0
932f0bc93bfc2cd857d4dcc4d53af04d
rooftop
true
afr_Latn
osm_description
[ "way/326644244" ]
australia
oceania
-42.887159
147.320153
null
building=yes
2.249858
https://www.openstreetmap.org/way/326644244
ODbL-1.0
f74db9f0846582f6adcae9d7f38b8619
rooftop
true
afr_Latn
osm_description
[ "way/326644339" ]
australia
oceania
-42.887677
147.313517
null
building=yes
105.319924
https://www.openstreetmap.org/way/326644339
ODbL-1.0
9fe0f67596a720bbc56cf2b37a17ba4b
rooftop
true
afr_Latn
osm_description
[ "way/326643848" ]
australia
oceania
-42.886876
147.314257
null
building=yes
138.270446
https://www.openstreetmap.org/way/326643848
ODbL-1.0
95c04d56930c0fa0e0c1a6224f19d5a7
rooftop
true
afr_Latn
osm_description
[ "way/326644632" ]
australia
oceania
-42.891526
147.319473
null
building=yes
58.138697
https://www.openstreetmap.org/way/326644632
ODbL-1.0
06e753f1df2281cc0bfd88d451ca52e8
rooftop
true
afr_Latn
osm_description
[ "way/326644552" ]
australia
oceania
-42.88662
147.313167
null
building=yes
81.279784
https://www.openstreetmap.org/way/326644552
ODbL-1.0
8b4e995e8da38bf74a32141a87f8c3cf
rooftop
true
afr_Latn
osm_description
[ "way/326644435" ]
australia
oceania
-42.886418
147.320241
null
building=yes
2.930541
https://www.openstreetmap.org/way/326644435
ODbL-1.0
d5120766fc16e0ea4d10cbf9d69f9bb4
rooftop
true
afr_Latn
osm_description
[ "way/326644742" ]
australia
oceania
-42.88982
147.315707
null
building=yes
131.330761
https://www.openstreetmap.org/way/326644742
ODbL-1.0
76b9fd3e37f4fbbe64b77602acffe535
rooftop
true
afr_Latn
osm_description
[ "way/326644142" ]
australia
oceania
-42.89023
147.318873
null
building=yes
152.032756
https://www.openstreetmap.org/way/326644142
ODbL-1.0
c6028f529e0fe1690aa4c3237cc154f3
rooftop
true
afr_Latn
osm_description
[ "way/326644608" ]
australia
oceania
-42.887243
147.32039
null
building=yes
108.177371
https://www.openstreetmap.org/way/326644608
ODbL-1.0
3908bee2f29fccceb7371d7206985e44
rooftop
true
afr_Latn
osm_description
[ "way/326643808" ]
australia
oceania
-42.887821
147.313595
null
building=yes
74.694759
https://www.openstreetmap.org/way/326643808
ODbL-1.0
83dc9b89795402737c76911da882880c
rooftop
true
afr_Latn
osm_description
[ "way/326644431" ]
australia
oceania
-42.886651
147.312863
null
building=yes
81.527636
https://www.openstreetmap.org/way/326644431
ODbL-1.0
a8589b5841fad0386257674a949acca3
rooftop
true
afr_Latn
osm_description
[ "way/326644562" ]
australia
oceania
-42.890482
147.316694
null
building=yes
149.543564
https://www.openstreetmap.org/way/326644562
ODbL-1.0
6d8a4c8efaf70843017b466373e915c9
rooftop
true
afr_Latn
osm_description
[ "way/326644510" ]
australia
oceania
-42.891074
147.314762
null
building=yes
34.836198
https://www.openstreetmap.org/way/326644510
ODbL-1.0
7fdbba37bccb14198cf7a40014063a36
rooftop
true
afr_Latn
osm_description
[ "way/326644430" ]
australia
oceania
-42.888055
147.319052
null
building=yes
21.530737
https://www.openstreetmap.org/way/326644430
ODbL-1.0
07d2426c4a45266fa1b480f78e36ad47
rooftop
true
afr_Latn
osm_description
[ "way/326644566" ]
australia
oceania
-42.889549
147.313273
null
building=yes
141.762238
https://www.openstreetmap.org/way/326644566
ODbL-1.0
0fd6155cc5c763367c1707cc75e1e93d
rooftop
true
afr_Latn
osm_description
[ "way/326644707" ]
australia
oceania
-42.888099
147.309979
null
building=yes
104.379291
https://www.openstreetmap.org/way/326644707
ODbL-1.0
0808a3c6aff8d37fcfe0965f0da0d342
rooftop
true
afr_Latn
osm_description
[ "way/326644471" ]
australia
oceania
-42.88662
147.318221
null
building=yes
9.617942
https://www.openstreetmap.org/way/326644471
ODbL-1.0
f98c4d3d0992368b1633bc5a8fd9014e
rooftop
true
afr_Latn
osm_description
[ "way/326644676" ]
australia
oceania
-42.887672
147.315872
null
building=yes
89.577652
https://www.openstreetmap.org/way/326644676
ODbL-1.0
689cac93a5f8f5e1480a351fff9fd042
rooftop
true
afr_Latn
osm_description
[ "way/326644527" ]
australia
oceania
-42.886776
147.318659
null
building=yes
6.048906
https://www.openstreetmap.org/way/326644527
ODbL-1.0
d8cfae7f09b38c8cf3bb909d8585e659
rooftop
true
afr_Latn
osm_description
[ "way/326644177" ]
australia
oceania
-42.89037
147.310381
null
building=yes
81.79183
https://www.openstreetmap.org/way/326644177
ODbL-1.0
138308892bc5bcbb397753946fd1a9f7
rooftop
true
afr_Latn
osm_description
[ "way/326644442" ]
australia
oceania
-42.890578
147.312021
null
building=yes
148.012594
https://www.openstreetmap.org/way/326644442
ODbL-1.0
f90382a9ea00f8ffd2069ea533b5f45b
rooftop
true
afr_Latn
osm_description
[ "way/326643745" ]
australia
oceania
-42.890277
147.315558
null
building=yes
123.981893
https://www.openstreetmap.org/way/326643745
ODbL-1.0
286cef4c751eb1cc3235efcdb325800c
rooftop
true
afr_Latn
osm_description
[ "way/326643750" ]
australia
oceania
-42.889149
147.310471
null
building=yes
159.528953
https://www.openstreetmap.org/way/326643750
ODbL-1.0
42a7de5810d40f9793f007971d7e9c92
rooftop
true
afr_Latn
osm_description
[ "way/326644071" ]
australia
oceania
-42.891525
147.319358
null
building=yes
232.207313
https://www.openstreetmap.org/way/326644071
ODbL-1.0
14e6431d9fb032cbcffb5aa859233047
rooftop
true
afr_Latn
osm_description
[ "way/326644253" ]
australia
oceania
-42.887807
147.315123
null
building=yes
12.191318
https://www.openstreetmap.org/way/326644253
ODbL-1.0
79e792bb592cf688d2464a4a4bf7c1e6
rooftop
true
afr_Latn
osm_description
[ "way/326644598" ]
australia
oceania
-42.887064
147.320027
null
building=yes
83.919348
https://www.openstreetmap.org/way/326644598
ODbL-1.0
0227d83a9b20303dad2e61f33e22996a
rooftop
true
afr_Latn
osm_description
[ "way/326644311" ]
australia
oceania
-42.890636
147.319831
null
building=yes
12.583088
https://www.openstreetmap.org/way/326644311
ODbL-1.0
9282182b7c7eed4dcdf196d8e9ce3ace
rooftop
true
afr_Latn
osm_description
[ "way/326644540" ]
australia
oceania
-42.887148
147.313832
null
building=yes
63.29552
https://www.openstreetmap.org/way/326644540
ODbL-1.0
4d1679938edb8cdda50fd40c0482639e
rooftop
true
afr_Latn
osm_description
[ "way/326644030" ]
australia
oceania
-42.890061
147.314193
null
building=yes
118.974222
https://www.openstreetmap.org/way/326644030
ODbL-1.0
1de1daec9fd421c6a1ca4463e6ecbc51
rooftop
true
afr_Latn
osm_description
[ "way/326644705" ]
australia
oceania
-42.890319
147.313219
null
building=yes
201.945871
https://www.openstreetmap.org/way/326644705
ODbL-1.0
a4b7b643336e40ac96c839f1aa1e1688
rooftop
true
afr_Latn
osm_description
[ "way/326643977" ]
australia
oceania
-42.888331
147.315935
null
building=yes
160.457936
https://www.openstreetmap.org/way/326643977
ODbL-1.0
1860eac1713d4c4bf4bf028d7f085ce4
rooftop
true
afr_Latn
osm_description
[ "way/326644200" ]
australia
oceania
-42.888151
147.315615
null
building=yes
2.834442
https://www.openstreetmap.org/way/326644200
ODbL-1.0
8e11c236c2f4b86153533eb5f8165216
rooftop
true
afr_Latn
osm_description
[ "way/326644085" ]
australia
oceania
-42.887498
147.320062
null
building=yes
86.920985
https://www.openstreetmap.org/way/326644085
ODbL-1.0
023230bce804de0fa9f1db05d32fcddf
rooftop
true
afr_Latn
osm_description
[ "way/326644006" ]
australia
oceania
-42.88855
147.314964
null
building=yes
136.15838
https://www.openstreetmap.org/way/326644006
ODbL-1.0
3043cec2a94afca95e645a0faaa520e7
rooftop
true
afr_Latn
osm_description
[ "way/326643831" ]
australia
oceania
-42.888609
147.314525
null
building=yes
196.956263
https://www.openstreetmap.org/way/326643831
ODbL-1.0
371820fb8343c3695de642daa79e5e57
rooftop
true
afr_Latn
osm_description
[ "way/326644662" ]
australia
oceania
-42.886812
147.31885
null
building=yes
88.725117
https://www.openstreetmap.org/way/326644662
ODbL-1.0
2d9754f2405fbc924dd4e09a679e4caa
rooftop
true
afr_Latn
osm_description
[ "way/326644738" ]
australia
oceania
-42.887827
147.315349
null
building=yes
104.330488
https://www.openstreetmap.org/way/326644738
ODbL-1.0
b4267886e0b6d8ab884a054ea1358cf0
rooftop
true
afr_Latn
osm_description
[ "way/326643774" ]
australia
oceania
-42.886412
147.317149
null
building=yes
110.677931
https://www.openstreetmap.org/way/326643774
ODbL-1.0
54994210427abeb0a251bd644fa15d64
rooftop
true
afr_Latn
osm_description
[ "way/326644151" ]
australia
oceania
-42.887061
147.311884
null
building=yes
168.464043
https://www.openstreetmap.org/way/326644151
ODbL-1.0
52069101947ae08f6a46c787d97f3446
rooftop
true
afr_Latn
osm_description
[ "way/326643785" ]
australia
oceania
-42.889956
147.311556
null
building=yes
215.915816
https://www.openstreetmap.org/way/326643785
ODbL-1.0
f80ebcb04b3a3278f9ba4456be9b8029
rooftop
true
afr_Latn
osm_description
[ "way/326644192" ]
australia
oceania
-42.887964
147.317175
null
building=yes
877.058494
https://www.openstreetmap.org/way/326644192
ODbL-1.0
9c400429438f6f343a1ecd294a30a965
rooftop
true
afr_Latn
osm_description
[ "way/326644368" ]
australia
oceania
-42.887617
147.320761
null
building=yes
21.633869
https://www.openstreetmap.org/way/326644368
ODbL-1.0
a2a356174fae5549455968c28f04889a
rooftop
true
afr_Latn
osm_description
[ "way/326644746" ]
australia
oceania
-42.89032
147.319013
null
building=yes
204.601776
https://www.openstreetmap.org/way/326644746
ODbL-1.0
29c517a2b0e65b62fe8c1a5019340593
rooftop
true
afr_Latn
osm_description
[ "way/326644449" ]
australia
oceania
-42.887211
147.313838
null
building=yes
4.766564
https://www.openstreetmap.org/way/326644449
ODbL-1.0
9359bb7975b07c67aa0dd31e75845bb8
rooftop
true
afr_Latn
osm_description
[ "way/326643961" ]
australia
oceania
-42.887105
147.312623
null
building=yes
140.424177
https://www.openstreetmap.org/way/326643961
ODbL-1.0
af897f478e6d03031fcd67deb74f049d
rooftop
true
afr_Latn
osm_description
[ "way/326644574" ]
australia
oceania
-42.890054
147.314972
null
building=yes
78.634377
https://www.openstreetmap.org/way/326644574
ODbL-1.0
9def0928e2ac27495b468ffda9293ace
rooftop
true
afr_Latn
osm_description
[ "way/326644634" ]
australia
oceania
-42.888689
147.314295
null
building=yes
268.120783
https://www.openstreetmap.org/way/326644634
ODbL-1.0
10a55734912839f2b4b9e845406ef521
rooftop
true
afr_Latn
osm_description
[ "way/326644658" ]
australia
oceania
-42.889949
147.318476
null
building=yes
116.545165
https://www.openstreetmap.org/way/326644658
ODbL-1.0
78bd572a33cfcecd27543e2b6a2bf132
rooftop
true
afr_Latn
osm_description
[ "way/326644708" ]
australia
oceania
-42.890672
147.31572
null
building=yes
124.420045
https://www.openstreetmap.org/way/326644708
ODbL-1.0
bcd136a7f75e054d56ed4dc755b223ff
rooftop
true
afr_Latn
osm_description
[ "way/326645314" ]
australia
oceania
-42.891782
147.317622
null
building=shed
5.227269
https://www.openstreetmap.org/way/326645314
ODbL-1.0
95d80af985dcf08b006a130a5a5c7c08
rooftop
true
afr_Latn
osm_description
[ "way/326644165" ]
australia
oceania
-42.886973
147.320645
null
building=yes
148.853386
https://www.openstreetmap.org/way/326644165
ODbL-1.0
4a2e2ff47105dd9ab9e6e398d365820f
rooftop
true
afr_Latn
osm_description
[ "way/326643810" ]
australia
oceania
-42.891649
147.319091
null
building=yes
116.257736
https://www.openstreetmap.org/way/326643810
ODbL-1.0
f31aa9c7b31953fa0dcc1e536677bd40
rooftop
true
afr_Latn
osm_description
[ "way/326644623" ]
australia
oceania
-42.888943
147.314008
null
building=yes
135.947892
https://www.openstreetmap.org/way/326644623
ODbL-1.0
b58fecfe2235133b7128992fa37ae911
rooftop
true
afr_Latn
osm_description
[ "way/326644232" ]
australia
oceania
-42.886448
147.31089
null
building=yes
12.109725
https://www.openstreetmap.org/way/326644232
ODbL-1.0
End of preview. Expand in Data Studio

GeoReSeT Fused: texts linked to OpenStreetMap polygons are filtered by land-use relevance; geotagged photos are clustered and described by a vision-language model; both are tied back to places

GeoReSeT Corpus Fused

96,675,303 sentences in 95 languages, tied to 3,704,364 OpenStreetMap polygons in 386 regions, each labelled for land-use / land-cover relevance by an LLM.

Plus 39,998 English scene descriptions, each labelled with a land-use / land-cover class by a vision-language model, tied to the locations of the 7,237,761 geotagged photos they describe.

Sentences come from the websites, description tags and Wikipedia / Wikivoyage articles linked to OpenStreetMap polygons. Every row is one sentence with a boolean landuse label, the language, and the location of the polygon it describes. Polygon shapes are in a separate polygons table. Scene descriptions come from ground-level photos (Mapillary, Flickr, Google Landmarks v2, iNaturalist, KartaView, camera traps) grouped by look with geo-rag; see Scenes. No photo is included.

Explore it on a map: the GeoReSeT Explorer shows the land-use sentences and scene photos on H3 cells, with their sources, languages and what places talk about.

The full dataset, all sentences and all scene photos (each photo stands for its scene's description at that place):

The full dataset per H3 cell

Quick start

from datasets import load_dataset

ds = load_dataset("NoeFlandre/GeoReSeT-Corpus-Fused", "fra_Latn", split="train")  # one language
ds = load_dataset("NoeFlandre/GeoReSeT-Corpus-Fused", "wikipedia", split="train")  # one source
ds = load_dataset("NoeFlandre/GeoReSeT-Corpus-Fused", split="train")  # everything
afg = load_dataset(
    "NoeFlandre/GeoReSeT-Corpus-Fused", "wikipedia", split="train", filters=[("region", "==", "afghanistan")]
)  # one region

Join sentences to polygon shapes with DuckDB, reading only what the query needs:

SELECT s.text, s.landuse, p.osm_tag, p.geometry
FROM 'hf://datasets/NoeFlandre/GeoReSeT-Corpus-Fused/data/wikipedia/*/*.parquet' s
JOIN 'hf://datasets/NoeFlandre/GeoReSeT-Corpus-Fused/polygons/*.parquet' p ON p.polygon_id = s.polygon_ids[1]
WHERE s.region = 'afghanistan' AND s.landuse;

Every photo location with the description and label of its scene, or one row per description:

photos = load_dataset("NoeFlandre/GeoReSeT-Corpus-Fused", "scene_photos", split="train")  # text, lulc_label, lat, lon, ...
scenes = load_dataset("NoeFlandre/GeoReSeT-Corpus-Fused", "scenes", split="train")
local = scenes.filter(lambda s: s["spread_km_p90"] <= 100)  # descriptions of one area

Sentences

Sentences per source: website, osm_description, then wikipedia and wikivoyage together:

Website sentences per H3 cell

OSM description sentences per H3 cell

Wikipedia and Wikivoyage sentences per H3 cell

Sentences the LLM labelled relevant (landuse is true):

Sentences labelled relevant per H3 cell

What places talk about, from the relevant sentences: each is embedded with microsoft/harrier-oss-v1-0.6b and counts in the H3 cell of its location (resolution 5). Each cell takes the mean embedding of its texts, and each of the 25,023 cells of resolution 4 holding 20 sentences or more the mean of its child cells, so that every place counts once, however much is written there. Those means are projected on their first three principal components, shown as red, green and blue. Within that colour, a cell is a little darker or lighter, and its hue a little turned, by how it differs from its surroundings: a second PCA, of each cell's mean minus its region's (resolution 3, about 12,000 km²), whose two axes are listed last:

Mean embedding of the relevant sentences per H3 cell, mean per resolution-4 cell, first three principal components as red, green and blue

  • PC1 (red, 14.3% of the variance): Cebuano place stubs ↔ Russian village histories. The cell's most frequent language explains 83% of it. Mostly language: read with care.
  • PC2 (green, 7.2% of the variance): airports and stations ↔ lakes, wetlands and islands. The cell's most frequent language explains 25% of it.
  • PC3 (blue, 6.3% of the variance): Cebuano airport stubs ↔ Iberian church architecture. The cell's most frequent language explains 62% of it. Mostly language: read with care.
  • Shade (darker ↔ lighter than the surroundings, 10.2% of how cells differ from them): airport buildings ↔ lakes and nature reserves.
  • Hue (turned back ↔ forward than the surroundings, 8.5% of how cells differ from them): airports ↔ village churches and heritage.

Each axis on its own, blue below zero and orange above. The relevant sentences at both ends of each axis are in pca_axes.md.

First principal component of the relevant sentences per resolution-4 cell

Second principal component of the relevant sentences per resolution-4 cell

Third principal component of the relevant sentences per resolution-4 cell

What places talk about, from the relevant sentences in English: each is embedded with microsoft/harrier-oss-v1-0.6b and counts in the H3 cell of its location (resolution 5). Each cell takes the mean embedding of its texts, and each of the 13,745 cells of resolution 4 holding 20 sentences or more the mean of its child cells, so that every place counts once, however much is written there. Those means are projected on their first three principal components, shown as red, green and blue. Within that colour, a cell is a little darker or lighter, and its hue a little turned, by how it differs from its surroundings: a second PCA, of each cell's mean minus its region's (resolution 3, about 12,000 km²), whose two axes are listed last:

Mean embedding of the relevant sentences in English per H3 cell, mean per resolution-4 cell, first three principal components as red, green and blue

  • PC1 (red, 11.1% of the variance): historic city buildings ↔ lakes and outdoor recreation.
  • PC2 (green, 9.5% of the variance): residential amenities ↔ islands and geology.
  • PC3 (blue, 7.1% of the variance): airports and cargo hubs ↔ rural churches and woodland.
  • Shade (darker ↔ lighter than the surroundings, 11.9% of how cells differ from them): campus and city buildings ↔ lakes and wilderness.
  • Hue (turned back ↔ forward than the surroundings, 9.3% of how cells differ from them): airports and transport ↔ old stone churches.

Each axis on its own, blue below zero and orange above. The relevant sentences in English at both ends of each axis are in pca_axes_english.md.

First principal component of the relevant sentences in English per resolution-4 cell

Second principal component of the relevant sentences in English per resolution-4 cell

Third principal component of the relevant sentences in English per resolution-4 cell

Source Sentences Words Polygons Languages Regions landuse share
wikipedia 58,052,517 933,468,335 678,711 85 386 40.4%
website 36,379,716 318,284,938 1,156,518 94 382 19.7%
wikivoyage 1,280,056 19,195,698 5,801 26 297 48.6%
osm_description 963,014 5,093,281 822,892 79 384 75.3%
total 96,675,303 1,276,042,252 2,613,738 95 386 33.0%

polygons counts distinct polygons linked to the sentences; the polygons table holds all 3,704,364. Words are whitespace-separated tokens, so Chinese, Japanese and Thai are undercounted.

Sentences per language, top 20

Continent Sentences Polygons
europe 55,158,209 1,449,445
north-america 20,336,824 555,832
asia 13,165,055 228,439
south-america 2,955,273 189,345
oceania 2,346,253 57,665
africa 2,143,242 121,316
central-america 495,980 10,859
antarctica 74,467 948

Polygon areas, over all 3,704,364 polygons:

10% 25% Median 75% 90%
84 m² 239 m² 917 m² 1.0 ha 8.5 ha

Fields

Field Type Description
id string Stable id of the sentence (128-bit hex)
text string The sentence, byte-exact as labelled
landuse bool The sentence describes land use, land cover or a geographic environment observable by remote sensing (LLM label)
language string ISO 639-3 + script (fra_Latn), see Languages
source string website, osm_description, wikipedia or wikivoyage
polygon_ids list[string] All polygons the sentence's document is linked to (way/123, relation/45)
region string Geofabrik extract of the first polygon (afghanistan, us-texas)
continent string Continent of that region
lat, lon float Centroid of the first polygon (WGS 84)
name string OSM name of the first polygon
osm_tag string Primary OSM tag of the first polygon (landuse=forest)
area_m2 float Area of the first polygon, m²
url string Website, OSM object, or Wikipedia / Wikivoyage revision permalink
license string License of the text: ODbL-1.0, CC-BY-SA-4.0 or other

The polygons config holds one row per polygon: polygon_id, osm_type, osm_id, name, osm_tag, region, continent, lat, lon, area_m2 and geometry (WKB, GeoParquet 1.1, WGS 84).

Languages

Labels follow GlotLID: ISO 639-3 + ISO 15924 script. How the language was found depends on the source:

  • website: GlotLID v3 on the whole page, so a sentence carries its page's label (an English menu line on a Hindi page is hin_Deva).
  • osm_description: Lingua, with GlotLID v3 as fallback.
  • wikipedia, wikivoyage: the language edition of the article (fr.wikipedia.org → fra_Latn), not detected.

Only text in the 85 languages of the SaT sentence splitter was split and labelled. Labels are finer than that: individual languages of a macrolanguage (nob_Latn and nno_Latn for Norwegian, cmn_Hani and yue_Hani for Chinese) and scripts (srp_Cyrl and srp_Latn) stay distinct. Other languages appear only in unlabelled, as unsplit.

Unlabelled sentences

The unlabelled config holds sentences without a usable label, kept to be labelled later. It has the same fields, with status and failure_reason instead of landuse:

  • failed: the model gave no usable yes/no answer (failure_reason, nearly always truncated).
  • unsplit: text the upstream splitter could not segment (unsupported language); text is a whole page, description or paragraph.
Source failed unsplit
osm_description 11,925 66,974
website 356,214 31,640
wikipedia 1,276,862 1,229,404
wikivoyage 29,787 144
total 1,674,788 1,328,162

Scenes

39,998 descriptions (4,061,169 words) of 7,237,761 geotagged ground-level photos in 250 countries, from geo-rag by Ananthu Aniraj. Photos were grouped by look into clusters (median 174 photos); a vision-language model described 4 representative photos of each cluster and labelled the scene with one of 40 land-use / land-cover classes.

Photos of the scenes per H3 cell:

Scene photos per H3 cell

What places talk about, from the scene descriptions: each is embedded with microsoft/harrier-oss-v1-0.6b and counts once in every H3 cell where its scene has photos (resolution 5). Each cell takes the mean embedding of its texts, and each of the 30,906 cells of resolution 4 holding 10 descriptions or more the mean of its child cells, so that every place counts once, however much is written there. Those means are projected on their first three principal components, shown as red, green and blue. Within that colour, a cell is a little darker or lighter, and its hue a little turned, by how it differs from its surroundings: a second PCA, of each cell's mean minus its region's (resolution 3, about 12,000 km²), whose two axes are listed last:

Mean embedding of the scene descriptions per H3 cell, mean per resolution-4 cell, first three principal components as red, green and blue

  • PC1 (red, 24.3% of the variance): paved urban streets ↔ arid rugged land.
  • PC2 (green, 18.0% of the variance): dry open land ↔ lakes and forests.
  • PC3 (blue, 10.3% of the variance): lush tropical vegetation ↔ polar ice and cold seas.
  • Shade (darker ↔ lighter than the surroundings, 31.4% of how cells differ from them): major paved roads ↔ greenery and wetlands.
  • Hue (turned back ↔ forward than the surroundings, 14.3% of how cells differ from them): dirt roads in open country ↔ dense historic built-up areas.

Each axis on its own, blue below zero and orange above. The scene descriptions at both ends of each axis are in pca_axes_scenes.md.

First principal component of the scene descriptions per resolution-4 cell

Second principal component of the scene descriptions per resolution-4 cell

Third principal component of the scene descriptions per resolution-4 cell

A description belongs to all the photos of its cluster, and a cluster groups photos that look alike, not photos of one place: half the clusters have photos in 20 countries or more. So:

  • scene_photos repeats the text and label of the scene on each of its photos, with the photo's own location, time and climate: one row per (text, place).
  • scenes holds each text once, with how far its photos are spread. Filter on spread_km_p90 for descriptions of one area.

Scenes whose photos lie, for 90% of them, within a distance of their median location:

10 km 100 km 1,000 km
967 3,327 9,433

Most frequent labels:

Label Scenes Photos
Transportation network 9,467 1,805,710
Historical / Cultural monument / Archaeological site 3,151 560,646
Natural grassland 2,833 501,638
Arid shrubland 2,571 451,462
Sparsely wooded / Savanna 2,374 438,627
High-density built-up 2,325 422,871
Suburban / Low-density residential 2,278 412,654
Industrial / Commercial zone 1,353 220,070
Temperate shrubland / Scrub 1,292 216,154
Broadleaved forest 1,105 206,233

Photos per platform, and the share with a public url:

Platform Photos With url
mapillary 3,684,627 100.0%
flickr 2,546,251 16.1%
wikimedia 836,052 0.0%
inaturalist 106,206 100.0%
kartaview 50,459 0.0%
snapshotusa 8,271 0.0%
wildobs 4,542 0.0%
wildlife_insights 696 0.0%
iwildcam 657 0.0%
Field (scene_photos) Type Description
scene_id int Scene (cluster) of the photo
text string Description of the scene (English, generated)
lulc_label string Land-use / land-cover class of the scene (generated)
lat, lon float Location of the photo (WGS 84)
h3_cell string H3 cell of the photo, resolution 11
country, continent string Where the photo lies (Natural Earth borders)
captured_at string Capture time (UTC, ISO 8601)
season, time_of_day string Climate-aware season (Wet Season, Summer, ...) and local time slot (Dawn, ...)
koppen string Köppen-Geiger climate zone (Cfb)
platform, photo_id string Where the photo comes from, and its id there
url string Public link to the photo; empty when its license forbids reuse or it is not online
photo_license string License of the photo, as given by its platform

The scenes config has one row per description: scene_id, text, visual_description (the model's first-pass description of the photos, without location context), lulc_label, parent_scene_id and parent_lulc_label (one of 2,000 coarser clusters), n_photos, lat / lon (median of its photos), spread_km_p50 / spread_km_p90 (distance from that median within which 50% / 90% of its photos lie) and countries (most frequent first).

Files

data/<source>/<language>/train-*.parquet   labelled sentences, sorted by region
unlabelled/train-*.parquet                 sentences without a label (see above)
polygons/train-*.parquet                   polygon shapes (GeoParquet)
scenes/train-*.parquet                     scene descriptions, one row per scene
scene_photos/train-*.parquet               one row per photo, with its scene's text and label
embeddings/train-*.parquet                 relevant sentences, one vector each (see [Methodology](#methodology))
scene_embeddings/train-*.parquet           scene descriptions, one vector each (see [Methodology](#methodology))
photo_labels/train-*.parquet              one row per photo, with its EUNIS label (see [EUNIS labels](#eunis-labels))
polygons_labels_shares/train-*.parquet     one row per polygon, with its EUNIS label and class shares (see [EUNIS labels](#eunis-labels))

Configs: all (default), one per source, one per language, unlabelled, polygons, scenes, scene_photos, embeddings, scene_embeddings, photo_labels, polygons_labels_shares.

EUNIS labels

photo_labels gives each photo its EUNIS habitat class (eunis_code, eunis_label), read from the EUNIS wall-to-wall map at the photo's coordinates. polygons_labels_shares gives each polygon the share of each class over the map pixels whose centre lies inside it (eunis_code, eunis_shares, max_eunis_share_code, max_eunis_share_label). If no pixel centre lies inside, the pixel under the polygon's centroid is used (label_method = centroid; otherwise polygon), with shares [1.0]. unlabelled_reason gives why a polygon has no label. Labels are null outside the map's coverage and on nodata pixels.

Methodology

  1. Polygons. Closed ways and multipolygon relations from the 386 regional Geofabrik extracts of OpenStreetMap, assembled with osmium.
  2. Text. Three kinds of text attached to a polygon:
  3. Language. GlotLID v3 for websites, Lingua with GlotLID fallback for descriptions, the language edition for Wikipedia and Wikivoyage (see Languages).
  4. Sentences. Split with SaT sat-3l-sm in its 85 languages; text in other languages is left whole (unsplit).
  5. Labels. Each sentence is classified by LiquidAI/LFM2.5-2.6B: does it describe land use, land cover or a geographic environment observable by remote sensing? Greedy decoding with reasoning, up to 4,096 tokens, served with SGLang and the DSpark draft model, on about 4,500 GPU-hours of Grid'5000. The model and prompt were chosen on benchmark-llms-landuse-relevance (53 models, 25,500 labelled sentences in 85 languages), where this setup reaches macro F1 0.81 and balanced accuracy 0.77. Labels and raw model outputs, with reasoning (code):
  6. This dataset. One row per sentence of each document (a document linked to several polygons appears once); overlapping extracts resolved to the most specific one; language codes normalized; sentences with fewer than two letters dropped, and Wikipedia / Wikivoyage rows made only of bracketed edit links or citations ([ edit | edit source ], [Smith 1998]) removed. Join the raw outputs on sha256(text) = text_sha256. Code: https://github.com/NoeFlandre/GeoReSeT-Corpus
  7. Scenes (geo-rag @ 4252e04). Geotagged photos scraped from Mapillary, Flickr, iNaturalist and KartaView, plus Google Landmarks v2 and camera-trap archives; duplicates removed on H3 cells and GPS glitches filtered. Photos embedded with google/tipsv2-b14 (CLS) and clustered with FAISS k-means into 40,000 clusters under 2,000 parents. For each cluster, google/gemma-4-E4B-it described a strip of 4 diverse photos near its centre (visual_description), then, given that description and the photos' location, season and Köppen climate, chose a land-use / land-cover class and wrote the final description (text). Here: the 2 clusters whose labelling failed are dropped, 6 labels outside the class list mapped to the nearest one, local file paths removed, Mapillary keys turned into links, and links removed for photos whose license forbids reuse. Raw input: NoeFlandre/geo-rag-scenes-raw @ 28fcda6 (private).
  8. Embeddings of the relevant sentences. The 27,495,185 distinct sentences with landuse true (the first id keeps a repeated text) embedded with microsoft/harrier-oss-v1-0.6b (sentence-transformers, bf16, instruction Instruct: Identify the land use or land cover described\nQuery: , last-token pooling, L2-normalised), stored as float16 in embeddings with their id, lat and lon; join on id for the rest. For the maps, each counts in the H3 cell of its location (resolution 5), each cell takes the mean of its texts, each parent cell of resolution 4 the mean of its child cells (each child counts once), parents with fewer than 20 sentences are dropped, and PCA is fitted on the parents' means (each counts once). Maps use the Equal Earth projection.
  9. Embeddings of the scene descriptions. The 39,998 distinct scene descriptions (text of scenes, id its scene_id) embedded with microsoft/harrier-oss-v1-0.6b (sentence-transformers, bf16, instruction Instruct: Identify the land use or land cover described\nQuery: , last-token pooling, L2-normalised), stored as float16 in scene_embeddings with their id, lat and lon; join on id for the rest. For the maps, each counts once in every H3 cell where its scene has photos (resolution 5), each cell takes the mean of its texts, each parent cell of resolution 4 the mean of its child cells (each child counts once), parents with fewer than 10 descriptions are dropped, and PCA is fitted on the parents' means (each counts once). Maps use the Equal Earth projection.

License

  • OpenStreetMap data (osm_description text, polygons, names, tags): ODbL 1.0, © OpenStreetMap contributors.
  • Wikipedia and Wikivoyage text: CC BY-SA 4.0; url points to the exact revision for attribution.
  • Website text (license = other): © the respective site owners, included for research. To have content removed, open a discussion on this repository.
  • Scene descriptions (text, visual_description, lulc_label): generated by Gemma 4 (Apache 2.0), released under CC BY 4.0.
  • Scene photos: no photo is redistributed. Each row gives the photo's own license in photo_license (Mapillary and KartaView: CC BY-SA 4.0; iNaturalist: CC BY-NC 4.0; Flickr and Wikimedia Commons: per photo). Photos marked All Rights Reserved or Do not reuse keep only their id and location, with no url.
Downloads last month
1,507

Space using NoeFlandre/GeoReSeT-Corpus-Fused 1

Collection including NoeFlandre/GeoReSeT-Corpus-Fused