WASHINGTON — Microsoft will supply Synthetaic, a startup that used artificial intelligence to track the Chinese spy balloon that zigzagged across the U.S. earlier this year, with digital resources that executives said will fuel advancements in computer vision and imagery analysis.
Under the five-year partnership, announced Aug. 29, Synthetaic will have access to nearly 1 million hours of cloud-computing power. The Wisconsin-based company said it will buttress its Rapid Automatic Image Categorization software, or RAIC, which lets users mine vast collections of photos and other visuals for specific items. Additional terms of the agreement were not disclosed.
“Using AI to extract insights from image data is like building a fire: you need heat (AI algorithm), fuel (data), and oxygen (compute),” Corey Jaskolski, the Synthetaic founder and CEO, said in a statement. “This partnership allows us to combine Azure’s GPU compute and fast data storage running next to our algorithms, which enables our customers to process massive quantities of visual data into actionable insights and models in minutes.”
RAIC was built on the Azure cloud. Jaskolski in a May interview told C4ISRNET that Synthetaic had several government customers and was in talks with the Air Force “on an effort to use RAIC on things like MQ-9 drone data.”
“It’s image recognition, but it’s done differently,” he said at the time. “When we tracked the balloon, we had just come up with this new feature that we call ‘geospatial object detect mode.’ So instead of just finding an area of a map that looked like other things, we could find discrete objects in satellite data.”
Overhead imagery is a resource of increasing importance, with governments, analysts and hobbyists tapping in to monitor their respective surroundings. Photos were used to expose Russia’s materiel buildup ahead of the February 2022 invasion of Ukraine, and, more recently, have shown the scale of devastation in Eastern Europe.
Synthetaic previously partnered with Planet Labs, a provider of Earth-observation imagery.
“We’ll be working with Planet to run this across Planet data, going forward and backward,” Jaskolski said. “It’s actually a really cool use case of RAIC because Planet has a six-year archive going back across the Earth. And, with RAIC, now we can search that archive and find anything.”