The problem Kitware (Booth 1430) aims to solve can be described as “Too Much Information:” The ever-expanding array of location-aware cameras and sensors around the world generates more geographically-tagged data than people can parse and process in a reasonable timeframe.
At GEOINT 2019, Kitware is demonstrating its open-source tools for automatic image and video recognition and analysis, showing how they can aid in tasks such as identifying disinformation campaigns via digital forensics, creating 3D models from 2D source material, tracking people and vehicles in video of varying quality, and identifying and indexing potentially millions of images and videos shared on social media.
“We work from source (sensors) to solution (answers), successfully creating advances to drive technology development in areas such as image recognition, object detection and tracking, scene understanding and recognition,” business development representative Libby Rose wrote via e-mail. “Our strengths include learning from small training sets through unsupervised learning, transfer/domain adaptation, incorporating synthetic data, and leveraging user feedback for interactive learning.”
Founded in 1988 and based in Clifton Park, N.Y., Kitware focuses on the public sector and counts the Defense Advanced Research Project Agency, the Intelligence Advanced Research Projects Activity, the Air Force Research Laboratory, and the Office of Naval Research among its customers.
Headline Image: Kitware’s open-source software uses deep learning for object detection and classification in satellite imagery. Photo Credit: Kitware