08.03.23 - Artificial intelligence can help coordinate relief operations and reconstruction scoping

Artificial intelligence can help coordinate relief operations and reconstruction scoping after natural disasters.

Although still in the early stage of development, the open source project xView2 has been recently deployed to help on ground rescue missions after the 2023 earthquakes in Turkey and Syria. The project has been supported by the Pentagon’s Defense Innovation Unit and Carnegie Mellon University’s Software Engineering Institute since 2019.

XView2 processes satellite imagery with a deep learning algorithm dedicated to object detection ("semantic segmentation") to identify building and infrastructure da mages and assess its severity. The algorithm can produce standardized maps of affected areas in minutes whereas it used to take hours, days or weeks to established a similar level of assessment relying on eyewitnesses, calls and field surveys. This rapid diagnosis addresses multiple stages of disaster response, from immediate assistance in coordinating and prioritising rescue teams efforts to find survivors by identifying the most affected areas, to long-term reconstruction planning efforts.

Ritwik Gupta, the principal AI scientist at the Defense Innovation Unit and a researcher at Berkeley aims to improve the algorithm’s capability and maintain the code open source and free that anyone can operate it as "a public service". 

#deeplearning #artificialintelligence #naturaldisaster #imagery