How AI can really help in disaster response.


Marash, Turkey: Satellite images (left) from Earth imaging company Planet Labs PBC and xView2 (right) from UC Berkeley, Department of Defense Innovation and Microsoft.

This is an improvement over traditional disaster assessment systems where rescue and emergency responders rely on eyewitness reports and calls to quickly identify where help is needed. In some recent cases, fixed-wing aircraft like drones have flown over disaster areas with cameras and sensors to provide human-assessed data. Take days, if not longer. Different response organizations often have their own data catalogs, further slowing common responses, making it challenging to create a standardized, common picture of which areas need help. xView2 can create a joint map of the affected area in minutes, helping organizations coordinate and prioritize responses—saving time and lives.

The obstacles

Of course, this technology is not a panacea for disaster response. There are several big challenges with xView2 that currently consume most of Gupta’s research attention.

The first and most important is how dependent the model is on satellite images, which provide clear photos during the day, when there is no cloud cover and when the satellite is overhead. The first usable images from Turkey did not come until February 9, three days after the first earthquake. And there are very few satellite images taken in remote and economically developed areas – for example, along the Syrian border. To address this, Gupta is investigating new imaging techniques, such as artificial aperture radars, which create images using microwave pulses instead of light waves.

Second, the xView2 model is up to 85 or 90% accurate in its assessment of damage and severity, but because satellite imagery has an aerial view, it cannot accurately detect damage to the sides of buildings.

Ultimately, Gupta says, getting on-the-ground organizations to use and trust an AI solution was difficult. “First responders are very traditional,” he says. “When they start telling you about this fantastic AI model, it’s not even on the ground and it’s looking at pixels like 120 miles in space, you’re not going to believe it at all.”

What’s next?

xView2 supports multiple phases of disaster response, from immediately evacuating damaged areas to assessing where safe temporary shelters can go to compensate for long-term rebuilding. ABB said it hopes xView2 will “become very important in our arsenal of damage assessment tools” going forward at the World Bank.

Since the code is open source and the program is free, anyone can use it. And Gupta intends to keep it that way. “When companies come in and start saying that. We can do this for business“I hate that,” he says. “This should be a public service that works for everyone.” Gupta is working on a web app so that any user can run reviews. Currently, organizations reach out to xView2 researchers.



Source link

Related posts

Leave a Comment

6 − 1 =