China's first sub-meter building footprints derived by deep learning (part 1 of 2).

Download

Due to Zenodo's file size limitations, we are releasing different parts of CBF and GBD in different versions. See the below for specific information:

1. China's first sub-meter building footprints (CBF) derived by deep learning:

part1: version12 (v12), City 1 to 210, https://doi.org/10.5281/zenodo.10473278 (current version)

part2: version13 (v13), City 211 to 356, https://doi.org/10.5281/zenodo.10475803

      Building attributes:

id: Index number of the current building.

year: Year of construction retrieved from GISA.

height_mean: The average height of the building (computed from the pixels within the building footprint) obtained from CNBH (meters).

height_max: Maximum height of the building (based on the highest pixel value within the building footprint) obtained from CNBH (meters).

height_min: Minimum height of the building (based on the lowest pixel value within the building footprint) obtained from CNBH (meters).

miniDist: Shortest straight-line distance to another building.

dist_id: Index number of the building with the shortest straight-line distance to the current building.

area: Area of the current building (square meters).

perimeter: Perimeter of the current building (meters).

inurban_19: A value of 1 indicates that the building was situated in an urban area in 1990, while a value of 0 signifies that it was located in a rural area in 1990. This determination is made using GUB data.

inurban_1: A value of 1 indicates that the building was situated in an urban area in 1995, while a value of 0 signifies that it was located in a rural area in 1995. This determination is made using GUB data.

inurban_20: A value of 1 indicates that the building was situated in an urban area in 2000, while a value of 0 signifies that it was located in a rural area in 2000. This determination is made using GUB data.

inurban_2: A value of 1 indicates that the building was situated in an urban area in 2005, while a value of 0 signifies that it was located in a rural area in 2005. This determination is made using GUB data.

inurban_3: A value of 1 indicates that the building was situated in an urban area in 2010, while a value of 0 signifies that it was located in a rural area in 2010. This determination is made using GUB data.

inurban_4: A value of 1 indicates that the building was situated in an urban area in 2015, while a value of 0 signifies that it was located in a rural area in 2015. This determination is made using GUB data.

inurban_5: A value of 1 indicates that the building was situated in an urban area in 2020, while a value of 0 signifies that it was located in a rural area in 2020. This determination is made using GUB data.

2. Global Building Dataset (GBD):

This dataset comprises approximately 800,000 images(512*512) with diverse architectural styles worldwide. It can be served as training and test samples for building extraction in different regions globally. In order to enhance usability, we did not break the continuity of the image and published it in 1024*1024 size.

Version
description
link

v1
All labels. Images of Africa, Australia, and South America.
https://zenodo.org/records/10043352

v2
image of Asia (part 1 to 30 of 53).
https://zenodo.org/records/10456238

v3
image of Asia (part 31 to 53 of 53).
https://zenodo.org/records/10457368

v4
image of Europe (part 1 to 21 of 58).
https://zenodo.org/records/10458273

v5
image of Europe (part 21 to 42 of 58).
https://zenodo.org/records/10460868

v6
image of Europe (part 43 to 58 of 58).
https://zenodo.org/records/10462506

v7
image of North America (part 1 to 20 of 93).
https://zenodo.org/records/10463385

v8
image of North America (part 21 to 40 of 93).
https://zenodo.org/records/10465076

v9
image of North America (part 41 to 60 of 93).
https://zenodo.org/records/10466569

v10
image of North America (part 61 to 80 of 93).
https://zenodo.org/records/10467291

v11
image of North America (part 81 to 93 of 93).
https://zenodo.org/records/10471557