A dataset of 129.6 million computer-generated building footprint polygons for the United States, derived from satellite imagery.
Microsoft USBuildingFootprints is a large-scale open dataset of building outlines for the entire United States. It solves the problem of manually mapping millions of structures by using computer vision on satellite imagery to automatically generate accurate polygon footprints. This provides a foundational geospatial layer for mapping, urban planning, and analysis.
Geospatial developers, data scientists, urban planners, OpenStreetMap contributors, and anyone needing a comprehensive building footprint dataset for the United States.
Developers choose this dataset because it offers a massive, ready-to-use collection of building footprints with verified high quality, saving immense manual digitization effort. Its open license and integration potential with tools like OpenStreetMap's RapiD and AI-assisted Tasking Manager provide immediate utility.
Computer generated building footprints for the United States
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Includes 129,591,852 building footprints across all 50 states and D.C., providing a nationwide dataset in standard GeoJSON format, as detailed in the download links.
Achieves 98.5% precision and 92.4% recall, with metrics showing quality often matches or exceeds hand-drawn OpenStreetMap data, based on polygon evaluation in the FAQ.
Released under the ODbL license, allowing free use and modification for both commercial and non-commercial purposes, as stated in the license section.
Directly supported in tools like the AI-assisted Tasking Manager and Facebook's RapiD, facilitating OpenStreetMap contributions, as mentioned in the FAQ.
Only 73 million footprints in the focal region are from 2019-2020 imagery; the rest average from 2012, making parts of the dataset outdated for time-sensitive applications.
The README explicitly warns that data quality varies by geography (e.g., rural vs. urban), necessitating local inspections before critical use, which adds overhead.
Unzipped files range up to 3.35 GiB per state, requiring significant storage and processing power, as shown in the download table, which can be prohibitive for some setups.