Our system ingests and analyzes big geographic and numeric data, including satellite photos, maps, laser scans, and much more. We work with 2D and 3D data from public and private sources, and the more data we have, the more accurate our models become.

We apply machine learning (including "deep learning") and algorithmic analysis techniques to understand the structure of the real world as represented in the data. Geopipe's system builds a semantic model of the world, representing not only what it looks like, but what's in it, from the ground to trees to the locations of individual windows in building facades.

By processing the data within a distributed system running in the cloud, we can rapidly analyze data and reconstruct whole cities in a matter of hours. Because we generate these models ahead of time, and update them as often as we have new data, you can immediately download models cropped from existing cities through our web interface.

GB of input data

km² of world models