Dataset Overview

Source of Imagery

Our strategy is to reuse images from existing benchmark datasets as much as possible and manually annotate new land cover labels. We selected xBD, Inria, Open Cities AI, SpaceNet, Landcover.ai, AIRS, GeoNRW, and HTCD datasets. For countries and regions not covered by the existing datasets, aerial images publicly available in such countries or regions were collected to mitigate the regional gap, which is an issue in most of the existing benchmark datasets. The open data were downloaded from OpenAerialMap and geospatial agencies in Peru and Japan. The attribution of source data is summarized here.

Classes and Annotations

We provide annotations with eight classes: bareland, rangeland, developed space, road, tree, water, agriculture land, and building. Their color and proportion of pixels are summarized below. All the labeling was done manually, and it took 2.5 hours per image on average.

Color (HEX)	Class	%
800000	Bareland	1.5
00FF24	Rangeland	22.9
949494	Developed space	16.1
FFFFFF	Road	6.7
226126	Tree	20.2
0045FF	Water	3.3
4BB549	Agriculture land	13.7
DE1F07	Building	15.6

Comparison with Related Datasets

OpenEarthMap presents a major advance over existing data with respect to geographic diversity and annotation quality (e.g., spatial details) as summarized below.

Summary of remote sensing benchmark datasets for semantic segmentation. B: building extraction, R: road extraction, LC: land cover mapping, and CD: change detection. The number of segments was counted on available labels.
Image level	GSD (m)	Dataset	Task	Classes	Countries	Regions	Area (km²)	Segments
Meter level	10	OpenSentinelMap	LC	15	---	---	505,202	3,467,552
	3	DynamicEarthNet	LC/CD	7	---	75	707	897,855
Sub-meter level	0.3--0.5	SpaceNet 1/2	B	2	5	5	5,555	685,235
	0.5/0.3/0.5	DeepGlobe	R/B/LC	2/2/7	---	---	2,220/984/1,717	---/302,701/20,697
	0.02--0.2	Open Cities AI	B	2	8	11	419	792,484
	0.5	xBD	B/CD	2/4	15	21	3,382	850,736
	0.3	LoveDA	LC	7	1	3	536	166,768
	0.25--0.5	OpenEarthMap	LC	8	44	97	799	2,205,395

License of Labels in OpenEarthMap

Label data of OpenEarthMap are provided under the same license as the original RGB images, which varies with each source dataset. For more details, please see the attribution of source data here. Label data for regions where the original RGB images are in the public domain or where the license is not explicitly stated are licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.