Thursday, August 25, 2016

DigitalGlobe Launches Vast Public Satellite Images Database For AI Training

Westminster, Colorado-based DigitalGlobe says it is contributing a significant amount of data into a new, publicly accessible trove of satellite images, which are intended for use for developing machine learning and "deep learning" algorithms for remote sensing. According to DigitalGlobe, it has launched SpaceNet, which will offer up an epository of satellite imagery and labeled training data, as part of an effort being supported by CosmiQ Works, NVIDIA, and Amazon Web Services. The satellite imagery company said that, by releasing the satellite imagery, it hopes to advance research in technology to automate mapping and image analysis using remote sensing. DigitalGlobe said the high resolution satellite imagery will include such information as the satellite imagery and 200,000 curated buildling footprints for Rio de Janeiro, which it says will hopefully help create create new algorithms to automate the extraction of features like buildings in dense urban environments. The new data set will eventually have more than 60 million, labeled satellite images.