2025-11-15 –, Way
fAIr is an open AI-assisted mapping tool developed by HOT, the Humanitarian OpenStreetMap Team. Recently a new machine learning algorithm was added and OSM users can experiment with mapping buildings around the globe. How well does fAIr work and has the new addition improved this?
The idea of (semi)automatically and effortlessly mapping OSM features from satellite images by means of some AI tool is fascinating to some, and terrifying to others.
Whether you belong to the AI-skeptical or to the more accommodating side of this dispute, it will be of interest to hear the results of our research on the performance of fAIr, a burgeoning tool developed by HOT, the Humanitarian OpenStreetMap Team. The aim of fAIr is to propose a fully open source mapping workflow, from the satellite imagery to the output data, for suggesting features for OSM, currently buildings footprints. It consists of an online AI mapping platform that allows users to create their own local training dataset, train/fine-tune a pre-trained AI model, and then map into OSM with the assistance of their own local model.
The research investigates the performance of both AI computer vision models available in fAIr. One of these was already presented in a few occasions in 2024 (at SoTM scientific track in Nairobi and at Geomob in London). We build up from that research and we show (if), how and where the tool works, and where there is room for improvement.
Humanitarian mapping, open source, AI, academic research.
Affiliation:The Alan Turing Institute, Heigit
A maps passionate currently employed as a Research Fellow at the Alan Turing Institute, in London, in the Urban Analytics team. As part of this role, Anna has carried out academic research to investigate the accuracy of a computer vision tool developed by HOT (Humanitarian OpenStreetMap Team) to map buildings from satellite imagery, and is currently working on this topic in collaboration with HeiGit (Heidelberg Institute for Geoinformation Technology, Germany).