The Defense Innovation Unit has started assessing algorithms designed to help first responders conduct search and rescue operations after the onslaught of a natural disaster.The organization said Wednesday it selected the algorithms through the xView Challenge, which is a computer vision competition encouraging developers to build a reusable platform for artificial intelligence and machine learning computations.
The program also intended to accelerate the development of satellite computer vision for humanitarian assistance and disaster relief.
During the competition, participants tested their algorithms against a set of overhead imagery of various scenes worldwide and over a million bounding box annotations across 60 object classes.
The DIU selected five winning algorithms that demonstrated the ability to advance computer vision capabilities, reduce minimum resolution for detection, boost learning capacity and further identify fine-grained object classes.
The organization facilitated the competition for six months in coordination with the National Geospatial-Intelligence Agency.
The second xView challenge will possibly be administered in 2019 to address issues in robustness and post-disaster change detection.
