Toggle post-event imagery and AI damage layers per area. Tap a cell for details.
fAIr detects building footprints on pre-event imagery; an AI damage model compares pre- and post-event imagery. fAIr and Microsoft AI for Good Lab detections are combined per area, then checked by MapSwipe volunteers: 4 to 7 people vote on each area (Yes = damaged/destroyed, No = no damage, Not sure). At a 50% accept threshold each area becomes confirmed damage, no damage, or uncertain. The per-building outputs are advisory; confirmed areas are the screening unit.
MapSwipe validation projects: Caraballeda · La Guaira · Caracas. Full data on the HuggingFace dataset.