What Is a Neural Compute Stick and Why Would You Want One?
Building on technology acquired through its purchase of Movidius last year, Intel has launched a USB tool and software developer kit designed to help developers add deep-learning capabilities to a wide range of "edge" devices.
The Movidius Neural Compute Stick uses deep neural network processing and machine vision technology to "reduce barriers to developing, tuning and deploying AI [artificial intelligence] applications," Intel said in a statement yesterday. It called the device "the world's first USB-based deep learning inference kit and self-contained artificial intelligence accelerator."
Available through Intel Movidius distributor partners RS Components and Mouser, the Neural Compute Stick is priced at $79. Intel said it's "designed to help democratize the machine intelligence space, and accelerate an age of ubiquitous intelligence devices and systems."
Enabling Offline AI
The Neural Compute Stick is built with vision processing unit (VPU) technology developed by Movidius that's already being used in devices ranging from drones to video surveillance cameras. The Myriad 2 VPU combines large computing capabilities with lower power demands, more than 100 gigaflops of performance at a power consumption level of one watt, according to Remi El-Ouazzane, vice president and general manager of Movidius. "This enables a wide range of AI applications to be deployed offline," El-Ouazzane said in the statement.
Prior to its acquisition by Intel for a reported $400 million in September, Movidius had released its own Fathom Neural Compute Stick, to help developers deploy trained neural networks to a range of devices equipped with Myriad 2 VPUs.
"It's going to mean that very soon, consumers are going to be introduced to surprisingly smart applications and products," El-Ouazzane said in a statement at the time. "It means the same level of surprise and delight we saw at the beginning of the smartphone revolution; we're going to see again with the machine intelligence revolution."
Applications for Robots, Smart Homes, Surveillance
In its new incarnation from Intel Movidius, the Neural Compute Stick lets developers bring trained neural networks built on the Caffe framework to embedded apps running on a variety of Myriad 2 VPU-based devices. These could include robots, as well as smart home, surveillance, and security devices.
Among the companies already using Movidius' Myriad 2 VPU are Dahua Technology, a China-based multinational that's the world's second-largest provider of video surveillance products, and DJI, a drone manufacturer that recently came out with the Spark mini-drone. The VPU enables the Spark to understand a controller's hand and arm gestures from a distance, and allows Dahua's video surveillance cameras to recognize faces, count people, analyze behavior, and detect illegally parked cars.
Besides Movidius, Intel has acquired some half-dozen companies over the 18 months aimed at boosting its capabilities in machine learning, computer vision and autonomous driving technology, including Nervana Systems and MobilEye.
"We're on the cusp of big breakthroughs in artificial intelligence. In the years ahead, we'll see new types of autonomous machines with more advanced capabilities as we make progress on one of the most difficult challenges of AI: getting our devices not just to see, but also to think," El-Ouzzane said in a September blog post announcing the Movidius acquisition.
Image credit: Movidius™ Neural Compute Sticks by Intel Corporation.
Posted: 2017-07-22 @ 11:39am PT
Fascinating potential here. I suspect that the underlying chipsets and firmware could be OEM'd in devices for far less than the $79. Probably under $20.