AI-Powered Drone Will Follow You Around and Take Pictures
Startup Skydio is prepping a consumer drone that uses machine vision to keep track of you while avoiding obstacles.
With high-quality drones now available for just a few hundred dollars, many consumers and businesses are taking to the skies. But many are plummeting to Earth, too—search YouTube for “drone crash” and you get more than three million results.
Entrepreneur Adam Bry argues that the drones on the market today are missing a key component needed to make them useful—the intelligence to fly autonomously. At MIT Technology Review’s EmTech Digital conference in San Francisco on Tuesday, Bry showed new footage from a drone that his startup, Skydio, will launch this year.
“Even in the enterprise I think people are finding it’s harder than they thought to use these,” said Bry. “If you’re an expert operator you can do really incredible things, but for mainstream users we’re just not there yet.”
The company’s first product is something like a flying, robotic camera crew. Bry’s video showed how the aircraft could follow people as they climbed trees, rode mountain bikes, and played football. When people moved behind obstacles, such as low-hanging branches, the drone would move around them or pause at a safe distance.
Skydio’s drone can do that because it navigates by analyzing images from its onboard cameras, which capture a 360-degree view around it. Drones on the market today typically rely on a combination of GPS and human steering.
Skydio’s flight software tracks objects around it to figure out, to within centimeters, how it is moving in the world. It harnesses the deep-learning techniques used by companies like Google to power image search that helps identify and track people against their background.
Bry said it can use cues from a person’s appearance and movement to avoid confusing individuals near one another, but he admitted it isn’t perfect. “If a person would have trouble telling two people apart, it would, too,” he said.
Bry declined to show images of Skydio’s drone or say exactly when it will be released. Although the product will be aimed at consumers, he said the flight control software would prove to have much broader uses.
“Something like this will be one of the foundations on which most of the other cool things in the drone industry get built,” he said. “Our bet is there’s this transition to autonomous in consumer [first], and then in industrial and commercial.”
Bry previously helped establish Google’s drone delivery program and was named to MIT Technology Review’s list of 35 young innovators in 2016.
Keep up with the latest in deep learning at EmTech MIT.
Discover where tech, business, and culture converge.
September 11-14, 2018
MIT Media Lab