One of the newest projects out of Microsoft Garage, Microsoft’s eponymous incubator, leverages artificial intelligence (AI) to pinpoint items of note in satellite imagery.
Earth Lens — a new open source iPad app launched today — identifies, tracks, and analyzes objects in photos taken miles above the planet’s surface. The source code is available on GitHub, and the app’s compatible with iPad devices running iOS 9.0 or above.
“Extracting insights from imagery data is an expensive process that could take upwards of days, delaying the decision-making process in mission-critical situations,” Lainie Huston, program manager at Microsoft Garage, wrote in a blog post. “You [typically need] a dedicated team to comb through thousands of satellite images.”
To expedite the process, Vancouver-based interns participating in Microsoft’s Garage Internship Program built an app that can, with the aid of satellite data and computer vision, assist in disaster relief and environmental conservation.
Earth Lens was designed in just 16 weeks, doesn’t require an internet connection, and taps a trained machine learning model provided by Microsoft’s AI for Earth team. According to the GitHub project page, it takes advantage of Apple’s CoreML framework to run the model locally.
The interface was informed by representatives at the Red Cross, OceanMind, and FarmBeats and streamlines the process of identifying, classifying, counting, aggregating, and labeling objects. Users can toggle object classes on and off in just a few taps or view satellite images in Time Series mode, which charts changes, patterns, and trends in a data visualization chart.
“Our vision for Earth Lens was to use technology and AI to transform the way humanitarian work is conducted,” said Michelle Chen, program manager intern for Earth Lens. “The rise of automated image recognition has empowered humanitarian organizations to triage damage and prioritized areas that require immediate help.”
It’s not Project Garage’s first foray into AI-powered apps. Another recent experiment — Snips Insights, which was also made available in open source — leveraged algorithms to draw information from screenshots on Windows 10 PCs. A screenshot of a tailored suit might pull up information about where it can be purchased, for example.