Just after sunrise on Saturday, a satellite put its camera long in the city of Mandalay in Myanmar, not far from the epicenter of Seriéré of magnitude 7.7 Friday which has devastated the country of Southeast Asia second largest city.
The mission was to capture images which, combined with artificial intelligence technologyCould help rescue organizations to quickly assess the number of buildings that had collapsed or were strongly damaged and where aid needed to go.
At first, the approach of high -tech computer vision did not work.
“The biggest challenge in this particular case was the clouds,” said Juan Lavista Ferres, chief scientist of Microsoft data. “There is no way to see through the clouds with this technology.”
The clouds finally moved and it took a few more hours to another Planet Labs satellite based in San Francisco to capture the aerial photos and send them to Microsoft AI Philanthropic for the right laboratory. At that time, it was already around 11 p.m. Friday at Microsoft Headquarters in Redmond, Washington. A group of Microsoft workers was ready and awaited the data.
The AI for the right laboratory has already made this type of evaluation of damage assisted by the AI before, following Libya’s catastrophic floods in 2023 or This year’s fires in Los Angeles. But rather than relying on a standard AI computer vision model that could run all visual data, they had to create a specific personalized version in Mandalay.
“The earth is too different, the natural disasters are too different and the imagery we obtain from satellites is simply too different to work in all situations,” said Lavista Ferres. For example, he said, while fires spread in a fairly predictable way, “an earthquake touches the whole city” and it can be more difficult to know the next day, where help is necessary.
Once the AI analysis was completeIt showed 515 buildings in Mandalay with 80% to 100% damage and 1,524 others with between 20% and 80% damage. Who showed the Generalized severity of the disasterBut, just as important, it helps to identify specific locations for damage.
“This is essential information for teams in the field,” said Lavista Ferres.
Microsoft warned that it should be used as a preliminary guide and will require a field verification for a complete understanding “. But in the meantime, the technological society has shared the analysis with aid groups such as the Red Cross.
Planet Labs says that his satellites – he has 15 orbit around the earth – have now photographed about a dozen sites in Myanmar and Thailand from Friday earthquake.
