State of the Map 2025 — fAIrSwipe

Presented "fAIrSwipe" — a 30-minute solo talk at State of the Map 2025 in Manila — introducing a new pipeline that merges fAIr's AI-generated building predictions with MapSwipe's crowdsourced validation workflow. The project automates the full journey from AI prediction to validated OSM data: fAIr generates building footprint predictions from satellite imagery; MapSwipe projects are automatically created for volunteer validation of those predictions; validated features are then conflated with existing OSM data and uploaded back to OpenStreetMap in a controlled manner.

Validation outcomes also feed back into fAIr to continuously improve the AI models. This closed-loop approach addresses one of the key challenges in AI-assisted mapping: ensuring AI predictions are validated by human volunteers before entering OpenStreetMap.

Watch the talk on media.ccc.de