Thursday, January 12, 2017

DigitalGlobe Aims At Machine Learning With Expanded Data

Westminster, Colorado-based satellite imaging provider DigitalGlobe said this week that it has expanded the amount of data it is providing to machine learning researchers, adding additional satellite imagery through its SpaceNet Challenge. The company said it released a new "points of interest" data for Rio de Janeiro, which includes more than 120,000 individual points of interest. DigitalGlobe said the new data comes via the help of the U.S. National Geospatial-Intelligence Agency, which licensed the dataset produced by DigitalGlobe. DigitalGlobe also said it has selected the first winners in its first "SpaceNet Challenge", which allowed developers to try their hand at creating algorithms to extract building footprints from satellite imagery, for a chance at $35,000 in prizes. DigitalGlobe said it would make those algorithms available to others via open source. DigitalGlobe said it plans another upcoming competition using high resolution, 30 cm imagery from WorldView-3 at new locations. Satellite images are commonly used in the intelligence world to track new buildings, construction, and other features at remote sites.