USBuildingFootprints  by microsoft

Dataset of building footprints for the United States

created 7 years ago
2,169 stars

Top 21.2% on sourcepulse

GitHubView on GitHub
Project Summary

This repository provides a comprehensive dataset of computer-generated building footprints for the entire United States, aimed at researchers, developers, and organizations involved in geospatial analysis, urban planning, and disaster response. The dataset offers over 129 million building footprints in GeoJSON format, derived using advanced computer vision techniques, with a focus on improving the OpenStreetMap ecosystem.

How It Works

The building footprint generation involves a two-stage process: semantic segmentation using an EfficientNet backbone with a combination of supervised and unsupervised training, followed by a novel polygonization method that approximates pixel predictions into polygons by incorporating a priori building properties. This approach aims to achieve higher quality and more accurate polygon representations compared to traditional methods like Douglas-Peucker.

Quick Start & Requirements

  • Download: Data is available for download per state or district in GeoJSON format. Links and file sizes are provided in the README.
  • Dependencies: Requires standard geospatial libraries for processing GeoJSON files. No specific software installation is mandated by the repository itself, as it's a data release.
  • Resource Footprint: Total unzipped size exceeds 30GB for the entire US dataset.

Highlighted Details

  • Claims pixel-based recall/precision of 95.5%/94.0% and polygon evaluation metrics of 98.5% precision and 92.4% recall.
  • Features a specific focal area (73 million footprints) derived from 2019-2020 imagery, with the rest averaging around 2012 imagery vintage.
  • Data is integrated into tools like Facebook's RapiD and Microsoft's AI assisted Tasking Manager.
  • Estimated false positive ratio is less than 1%.

Maintenance & Community

This is a data release from Microsoft Maps. Contributions are welcomed via pull requests, subject to a Contributor License Agreement (CLA). The project adheres to the Microsoft Open Source Code of Conduct.

Licensing & Compatibility

Licensed under the Open Data Commons Open Database License (ODbL). This license permits free use, distribution, and modification, but requires attribution and sharing of derivatives under the same license. Commercial use is generally permitted, but users should review ODbL terms.

Limitations & Caveats

The quality of the data varies geographically and between urban/rural areas. Users are advised to inspect local quality and consult OpenStreetMap community guidelines before importing data to avoid overwriting existing contributions.

Health Check
Last commit

8 months ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
20 stars in the last 90 days

Explore Similar Projects

Feedback? Help us improve.