A new project by Carnegie Mellon University researchers provides journalists, citizen scientists, and other researchers with the ability to quickly scan large geographical regions for specific visual features.
Simply click on a feature in the satellite imagery – a baseball diamond, cul-de-sac, tennis court – and Terrapattern will find other things that look similar in the area and pinpoint them on the map.
Using a deep learning neural network trained for five days on an NVIDIA GeForce GPU, their model will look at small squares of the landscape and, comparing those patterns to a huge database of tagged map features from OpenStreetMap, it learned to associate them with certain concepts.
Currently, Terrapattern is limited to Pittsburgh, San Francisco, New York City and Detroit, but access to more cities is coming soon.
Read more >>
Open-Access Visual Search Tool for Satellite Imagery
May 26, 2016
Discuss (0)

Related resources
- GTC session: GPU-Accelerating End-to-End Geospatial Workflows (Spring 2023)
- GTC session: On-Board Satellite Component Recognition for Autonomous Missions in Orbit (Spring 2023)
- GTC session: cuSpatial: Integrate High-Performance Spatial Computation with Your Existing Workflow (Spring 2023)
- Webinar: Inception Workshop 101 - Getting Started with Vision AI
- Webinar: Inception Workshop 101 - Getting Started with Data Science
- Webinar: Omniverse for Digital Twins Building