Inside every Phantom 4 is Movidius' Myriad 2 chip, the two companies announced Wednesday. The chip is what Movidius calls a visual processing unit (or VPU). The low-powered chip is tailored for machine vision. All of the Phantom 4's cameras and sonar sensors are plugged into the chip that is then able to map out and understand the environment. The drone is able to rely almost entirely on visual cues to avoid running into things and stay in a fixed position.
"I met with [DJI CEO Frank Wang] about 18 months ago and he was talking to me about his dream of really trying to bring more autonomy into a drone,” said Movidius CEO Remi El-Ouazzane. “He wanted everyone to really enjoy the experience of flying drones without being a perfect pilot. His vision really resonated with us. We started to sketch out how to work together.”
With this partnership, Movidius has suddenly found itself as arguably one of the most important chip makers for drones. Large chip makers such as
"I truly believe DJI is at least a year ahead of everyone else when it comes to deploying this computer vision technology," El-Ouazzane said. "And Frank isn't going to stop there."
Founded back in 2005 in Ireland, Movidius has raised a total of $95 million in venture capital, according to private market research firm PitchBook. The single biggest chunk of money came last April with $40 million. To date, Movidius' only other publicly-announced partnership has been with Google. Google and Movidius launched a partnership in 2014 around Project Tango, Google's attempt at bringing 3D sensing to mobile devices. Early Project Tango prototypes of a phone and tablet used Movidius' Myriad chips. And earlier this year, Movidius also announced a partnership with Google's machine learning group to work on unannounced devices that put artificial intelligence into low-powered devices.
“We’re in the space of giving a visual cortex to devices,” El-Ouazzane said. “Where we’re at today is the vision capabilities of a one-year-old. The goal is to provide the technology to make devices completely autonomous. We have a long road ahead of us.”