ORNL has developed a new US Building Dataset using AI, focusing on accurate building footprints with metadata including addresses and building usage. The dataset was enriched with data from Lightbox, OpenStreetMap, and US government agencies. The author details their workstation specs and setup for analyzing the dataset. They share the process of downloading and analyzing the 131M US Building Dataset using tools like Python, DuckDB, and QGIS. Unique building information like UUIDs, heights, and occupational classes is highlighted, along with a breakdown of data by state, city, and building use. The author also addresses challenges such as erroneous data values and extensive metadata.
https://tech.marksblogg.com/ornl-fema-buildings.html