TAMPA, FLA. — The nation’s intelligence agencies have more data than they know what to do with. What they really need are software algorithms that can magically turn data into that elusive insight known as actionable intelligence.
Giving away data to software developers has been a proven tactic for industries trying to create excitement and new products. The geospatial intelligence world is continuing that trend in a big way. At its annual symposium this week, the United States Geospatial Intelligence Foundation presented its “Industry Achievement Award” to SpaceNet, an open repository of free imagery with map features created by CosmiQ Works, Radiant Solutions and NVIDIA.
“This enables the entire research and development community to focus on developing solutions to challenging geospatial problems instead of spending time on data acquisition,” said David Lindenbaum, principal engineer at CosmiQ Works. He believes SpaceNet will “lower the barrier to entry for developers and startups to access high-quality geospatial data.”
“It’s exciting,” Kevin Berce, senior director at NVIDIA, told SpaceNews. “This allows people to go find out what the problems are in using overhead imagery.”
Amid a frenzy over how the U.S. military and intelligence community will use machine learning and artificial intelligence, the industry is rushing to “teach machines” to do humanly impossible tasks like analyze a trove of data and instantly find specific objects or people. To avoid the garbage-in garbage-out problem, developers need data that is just as good as what companies sell at steep prices. Now they can get it for free, as long as they agree to the terms of service set by the providers.
SpaceNet, launched in 2016, is the first public release of multi-spectral satellite imagery of 50 cm resolution. It may be used for national security purposes but the industry hopes to promote other applications of machine learning using geospatial data such as humanitarian crisis response, ensuring the accuracy of maps used by self-driving cars, predicting earthquakes or even preventing genodice.
Developers are invited to participate in SpaceNet “challenges.” Three have been conducted so far. The fourth is coming up later this year and will focus on the identification of objects at high off-nadir angles of collection, as opposed to straight down views, to mimic imagery that may be collected by newer constellations with higher revisit rates. The data repository includes over 5,700 square kilometers of DigitalGlobe imagery, more than 685,000 building footprints, and more than 8,000 km of road networks hosted on Amazon Web Services.
The remote sensing industry is trying to contribute to “global transparency,” said Walter Scott, chief technology officer of Maxar Technologies and founder of DigitalGlobe.
“We are enabling nations to act on the basis of facts not fear,” he said in an interview. “It’s about shining a light on atrocities, making it harder for governments to harm their own people, or helping with vaccine deliveries,” Scott said. “Most people don’t appreciate just how badly mapped most of the world is.”
Scott pushed back on the idea that offering data to developers at no cost means everyone else will expect free data at the expense of companies like DigitalGlobe. “We want to see more apps developed,” he said. “It’s in everybody’s interest to make data available.” With the boom in machine learning, tools have to be made available to a much broader community of developers,” Scott said. “We are planting seeds.”
The motherlode of geospatial data, the National Geospatial Intelligence Agency, also is getting in the game. NGA Director Robert Cardillo on Monday announced the agency’s new project called GeoWorks. “It is a much easier way to access and work with NGA’s data,” he said. “Now any U.S. company or academic institution or interested individual can gain access to our data and tools, and build something.”
The Pentagon’s Silicon Valley technology outreach office, the Defense Innovation Unit Experimental, teamed up with NGA on a competition earlier this year where companies were given government satellite imagery to develop machine learning algorithms. The xView Detection Challenge was promoted as a way to “advance key frontiers in computer vision and develop new solutions for national security and disaster response.”
Robbie Schingler, co-founder and chief strategy officer of Planet Labs, said the industry wants to see “disruptive tools.” Machine learning has been around for many years, but “algorithms get better with more data and more training,” he told SpaceNews. “In the computer science industry we’ve been doing this for decades.” Data storage and computing power have gotten cheaper, and the industry has turned its attention to algorithms. “To get the accuracy you want, you need a lot of data. There’s no other place in the world that has more data than NGA,” Schingler said. “They know it’s a game changer for them to get more accurate algorithms.”
Data and platforms have been commoditized and “everybody now wants solutions,” said Robert Laudati, managing director of commercial products at Harris Space and Intelligence Systems. “Data is becoming free, easier to collect, and data vendors are moving into the analytical space to solve problems for customers.”
Artificial intelligence is everywhere, and the next step is “how we operationalize it,” Laudati told SpaceNews. “How do we stand up the systems to drive this massive amount fo data and do something about it?”
