The DIU which is a part of Defense Department technology outreach division based in Silicon Valley has posted images from satellite for its 2nd prize challenge referred to as xVIEW2 that focuses on usage of artificial intelligence for assessment of natural disasters related damage. The contestants for this competition are vying for prizes worth $150000 and they have to develop computer vision related algorithms that will automate process which will help in assessing damage to buildings and other man-made structures after natural disasters like tsunamis, floods, earthquakes, wildfires, wind and volcanic eruptions.
The images posted by DIU are high resolution pictures of 550,230 buildings from across 10 countries that were obtained by Maxar’s DGOD Program. The images from satellite shows building conditions before and after natural disasters, along with details of damage scores for every building and their annotated polygons. DIU stated that these pictures are one of the largest public datasets of high quality annotated high resolution satellite images. DIU’s AIP director Mike Kaul stated that the organization is keen on enlisting an international community of machine learning experts to manage a critical problem. This refers to detection of key objects within overhead imagery to assess damage during disasters.
The DIU stated that more than 3000 people have joined to compete and will be judged on how fast their algorithms are able to detect and access the spread of damage. The purpose is to help in automating the imagery analysis and helping government agencies to speed up relief measures to affected areas. The algorithms of contestants will be checked against a new dataset referred to as xBD that has been created by experts from academic field and industry. It is presently the most diverse and largest annotated dataset related to building damage that has been used for generating and testing models that are meant for automating assessment of building damage.