NASA-Backed GeoAI Package Lets Researchers Run Satellite Analysis With a Few Lines of Python

Key points

• GeoAI is a free, open-source Python package for applying artificial intelligence to satellite and geographic data, published in the Journal of Open Source Software in 2026.
• The project received partial funding from NASA (Grant No. 80NSSC22K1742) and the U.S. Geological Survey (Grant No. G23AP00683).
• It works with satellite sources including Sentinel and Landsat, and exports results in standard map formats such as GeoJSON and Shapefile.
• A no-code QGIS plugin lets users run AI mapping workflows inside a desktop map application without writing any code.
• The package is free to install via pip or conda and is licensed under the permissive MIT open-source licence.

Spotting a flooded field from a satellite image used to need a team of specialists and weeks of work. GeoAI wants to shrink that to an afternoon and a laptop.

What does GeoAI actually do?

It connects AI models to geographic data: satellite pictures, aerial photos, digital maps. Point it at a city and it can outline every building. Aim it at farmland and it can label which fields are crops or bare soil.

The package handles the tedious steps automatically, downloading imagery from providers like Sentinel and Landsat, slicing large images into tiles for training, and saving results in standard map file formats that any GIS (geographic information system, software used to create and edit maps) can open. For people who'd rather not write code at all, there's a plugin for QGIS, a free desktop mapping application. Install the plugin, click a few buttons, and the AI does the rest.

Who is paying for this, and why does that matter?

Government money is behind it. NASA contributed through its Open Source Tools programme, and the U.S. Geological Survey added further support through a five-year cooperative agreement running from 2023 to 2027. That funding matters because the project isn't trying to sell you anything. The code is free, the MIT licence lets anyone use it commercially, and the documentation sits at opengeoai.org at no charge.

NASA's appetite for open geospatial AI has been clear for a while. We reported in July that NASA ran a Google AI model inside a satellite, cutting wildfire-alert times from 90 minutes toward near-instant. GeoAI sits at the other end of that pipeline: the ground-level tooling that lets researchers do something useful with what satellites send back.

Author Qiusheng Wu also published a companion book, "GeoAI with Python: A Practical Guide to Open-Source Geospatial AI", available in PDF and print editions at book.opengeoai.org.

Should ordinary people care about this?

Directly, probably not yet. GeoAI is a researcher and developer tool. But the downstream effects are real. Environmental agencies tracking deforestation, city planners counting rooftops for housing surveys, disaster-response teams mapping flood damage: all of them could use tools built on software like this. Faster, cheaper satellite analysis means the reports that shape your community get made with better data.

For GIS professionals and data-minded scientists reading this, the barrier to running a building-detection model just dropped considerably. The package supports GPU acceleration (specialised chips that speed up AI calculations) automatically when available, so it scales from a laptop to a research cluster without changing your code.

If you work with geographic data professionally, install the package (pip install geoai) and run one of the tutorial notebooks at opengeoai.org before paying for a commercial solution. The government already funded the development.