Intel Unveils USB Toolkit for AI Prototyping
Intel has begun rolling out AI hardware previewed last fall with the release of its Movidius Neural Compute Stick, billed as a self-contained AI accelerator in a USB format that requires no network connection.
The chipmaker said Thursday (July 20) the device is intended to make deep learning applications more accessible to AI researchers and product developers to they can pursue offline AI prototyping. The USB toolkit compiles and accelerates “neural networks at the edge,” the company said.
Intel’s machine vision strategy focuses on combining CPUs and vision processors with cameras to allow devices to understand their surroundings.
The neural stick device, which includes on-board processing and does not require cloud connectivity, is targeted at AI applications for “end devices” such as robots and drones as well as object tracking and object recognition. Potential object recognition applications include vehicle backup cameras or collision avoidance apps like smart vacuum cleaners.
Intel (NASDAQ: INTC) also is pitching the neural compute stick for video analytics applications along with improving computer vision algorithms used in commercial drones for applications such as inventory inspection.
The compute stick, a standard USB 3.0 drive, is among a series of AI hardware implementations and development tools in Intel’s pipeline. The heart of the USB-based device is the Movidius Myriad 2 vision-processing chip capable of handling more than 100 gigaflops within a 1-watt power envelope. The ability to run real-time deep learning networks from the device “enables a wide range of AI applications to be deployed offline,” explained Remi El-Ouazzane, vice president and general manager of Movidius, the computer vision startup Intel acquired last September.
The device converts convolutional neural networks into an embedded neural network running atop the Myriad VPU. A tuning feature allows developers to validate scripts to compare accuracy of customized models to the original. The device can then be used as a neural network accelerator that adds deep learning inference capabilities, the company said.
Intel earlier acquired deep learning specialist Nervana Systems and cognitive computing developer Saffron. Those deals along with the Movidius acquisition are aimed at new machine vision capabilities targeting object recognition and scene understanding as well as tracking and navigation aids, according to Josh Walden, general manager of the chipmaker’s New Technology Group.
“Our goal is to compress the innovation cycle from conception to the deployment of increasingly intelligent, robust and collaborative machines,” Intel CEO Brian Krzanich noted during a company AI event last November.
The chip giant was scheduled to test the first AI-specific hardware, code-named “Lake Crest,” in the first half of 2017, with limited availability later in the year. Lake Crest will be optimized for running neural network workloads.
The neural compute stick will be available for purchase at the Computer Vision and Pattern Recognition event that begins on July 21. The price tag is $79.