Jane Edwards
Tony Frazier, senior vice president of government solutions at DigitalGlobe, has said the company works to analyze large amounts of satellite images and data through deep-learning tools and crowdsourcing, GCN reported Wednesday.
Frazier told GCN columnist Patrick Marshall in an interview that OpenCV computer vision software library, Caffe deep-learning framework and NVIDIA graphics processors are some of the deep-learning systems that DigitalGlobe uses to initially screen satellite images for objects and events.
He also noted that the use of deep-learning algorithms has helped the company narrow down the search for infrastructure, objects and activities for government customers.
Such algorithms that run on the DigitalGlobe-built Geospatial Big Data platform learn over time through the integration of human analyses of such images generated via crowdsourcing, Frazier added.
