fAIr¶
fAIr is an open, AI-assisted mapping service for humanitarian and development work, from the Humanitarian OpenStreetMap Team (HOT). It is a shared collection of geo-AI models: a mapper picks a model, runs it over their own area to detect features such as buildings or trees, reviews the results, and can adapt the model to local conditions. A mapper stays in control at each step.
The software is free and open source. Predictions are suggestions for a mapper to review. fAIr offers its outputs as open data layers, and bringing AI-generated features into OpenStreetMap is outside its aim; see fAIr and OpenStreetMap.
Download the fAIr explainer (PDF)
Where fAIr is today¶
fAIr is an open collection of models whose predictions are checked by people, in fAIr and through MapSwipe. Current figures are on Milestones, and the direction ahead on Vision.
Who it is for¶
-
Mapper
Picks a model and runs it on their own area, then validates the results.
-
Advanced user
Retrains a model for their area and publishes the local model.
-
Developer
Writes new geo-AI models and adds them to fAIr for others to use.
Start here¶
-
What is fAIr
What it is and why it exists.
-
How it works
The model loop and the data pipeline.
-
Models
What fAIr can map, and the open model catalog.
-
Using fAIr
Run a model on your own area and get the results.
-
Developer setup
Run the full stack locally and start iterating.
-
Datasets
Published training and prediction data.
-
Roadmap
History and what is planned next.