The Digital Press

A new dataset called GlobalBuildingAtlas (GBA), published by researchers at the Technical University of Munich, estimates there are about 2.75 billion buildings worldwide, challenging the UN’s estimate of 4 billion. The dataset includes detailed building footprints and heights derived from AI analysis of satellite imagery, and has been converted into efficient Parquet files for easier use.

Why it matters: GBA provides a massive, detailed global building dataset useful for urban planning, environmental studies, and mapping applications.

The big picture: AI-driven satellite imagery analysis enables comprehensive global datasets that were previously unavailable or incomplete, improving geospatial data accuracy.

Stunning stat: The dataset contains over 2.7 billion building records, surpassing other major sources like Overture’s 2.5 billion buildings.

Commenters say: They note the impressive scale and technical effort but raise questions about AI accuracy, data currency, environmental impacts like roof painting, and socioeconomic reflections on housing access.