Damage assessment is important in order to start search and rescue activities in natural disasters such as the earthquake in Turkey. The artificial intelligence system xView2 accelerates this process with deep learning.
We often hear big promises about the potential of artificial intelligence to solve problems in the world. Although most of them are not realistic, the artificial intelligence system xView2, developed with deep learning, can save lives in natural disasters such as the earthquake disaster in Turkey.
xView2, an open source project developed in 2019 by the Pentagon’s Defense Innovation Unit and Carnegie Mellon University’s Software Engineering Institute, was used by ground crews in search and rescue activities and damage assessment after the earthquake in Turkey. In the meantime, there is support from many institutions and organizations, including Microsoft, the University of California and Berkeley, in the development of xView2. xView2 combines satellite imagery with machine learning to identify building and infrastructure damage in the disaster area and quickly detect the severity of the damage.
In the past five years, xView2 has been used during forest fires and post-flood rescue efforts in Nepal. Ritwik Gupta, chief artificial intelligence scientist at the Defense Innovation Unit, explained in his statement that xView2 was used by at least two different ground crews in Adıyaman, which was devastated after the earthquake. It is underlined that the vehicle is also used in the detection of unnoticeable but damaged areas in the earthquake zone. “If we can save even one life, we’re putting this technology to good use,” Gupta said. uses expressions.
How xView2 Helps
The algorithms in xView2 use a pixel-based object identification technique on satellite images of the region. As you can see from the images, the darker the red color that represents the buildings, the worse the condition of the debris. Atishay Abbhi, a disaster risk management specialist at the World Bank, states that normally this damage assessment can take days, weeks, and sometimes months, but thanks to artificial intelligence, the process is completed in a few hours or minutes, depending on the scale.
With artificial intelligence systems like xView2, there may be less need for eyewitness reports and denunciations. On the other hand, although this artificial intelligence system saves a lot of time, it needs clear satellite images. Therefore, there are obstacles to the use of the system in cloudy weather. Second, while the xView2 model is up to 85% or 90% accurate in its precise assessment of damage and severity, satellite imagery cannot provide insight in detecting damage to the edges of buildings because it has an aerial perspective.