Login
- Title:
GlobalBuildingMap
- Document type:
- Forschungsdaten
- Publication date:
- 19.09.2025
- Responsible:
- Zhu, Xiaoxiang
- Authors:
- Zhu, Xiaoxiang; Li, Qingyu; Shi, Yilei; Wang, Yuanyuan; Stewart, Adam J.; Prexl, Jonathan
- Author affiliation:
- TUM
- Publisher:
- TUM
- Identifier:
- doi:10.14459/2024mp1764505.002
- Concept DOI:
- doi:10.14459/2024mp1764505
- End date of data production:
- 01.11.2024
- Subject area:
- DAT Datenverarbeitung, Informatik; GEO Geowissenschaften; UMW Umweltwissenschaften
- Other subject areas:
- Remote Sensing, Earth observation, machine learning
- Resource type:
- Experimente und Beobachtungen / experiments and observations; Abbildungen von Objekten / image of objects
- Data type:
- Bilder / images
- Description:
- This is the second version of the dataset. The GlobalBuildingMap (GBM) dataset provides the highest resolution and highest accuracy building footprint map on a global scale ever created. GBM was generated by training and applying modern deep neural networks on nearly 800,000 satellite images. The dataset is stored in 5 by 5 degree tiles in geotiff format. This is the second version of the dataset, which contains the whole world. Link to the first version: https://mediatum.ub.tum.de/1764505?v=1
- Method of data assessment:
- The GlobalBuildingMap is generated by applying an ensemble of deep neural networks on nearly 800,000 satellite images of about 3m resolution. The deep neural networks were trained with manually inspected training samples generated from OpenStreetMap. Evaluation of GlobalBuildingMap were conducted on 34 unseen cities. It reaches an overall F1 score of 0.54, comparing to 0.47 for Microsoft building footprint and 0.20 for Google building footprint.
- Links:
- Please note: The file name of the PlanetScope scenes used to produce this dataset can be found here https://github.com/zhu-xlab/GlobalBuildingMap/blob/main/assets/downloaded_items.txt
- Key words:
- global building footprint; PlanetScope; machine learning; deep learning; convolutional neural network; So2Sat
- Technical remarks:
- View and download (53,9 GB total, 939 Files)
The data server also offers downloads with FTP
The data server also offers downloads with rsync (password m1764505.002):
rsync rsync://m1764505.002@dataserv.ub.tum.de/m1764505.002/ - Language:
- en
- Rights:
- by, http://creativecommons.org/licenses/by/4.0
- Horizon 2020:
- ERC-2016-StG-714087
BibTeX
versions
