Movidius Launches Neural Compute Stick

The famous chipmaker Movidius has launched a new plug and play USB Neural Compute Stick that is able to add a machine vision processor to any device. The device contains the same processor that DJI autonomous drones encapsulate (provided by Movidius) but the Compute Stick is more user-friendly because of its portability and plug-n-play capability.
 
The company describes it as,
 
"Movidius Neural Compute Stick is a modular artificial intelligence (AI) accelerator in a standard USB 3.0 stick form factor. Designed for product developers, researchers and makers, Movidius Neural Compute Stick aims to reduce barriers to developing, tuning and deploying deep learning applications at the edge by delivering dedicated high-performance deep neural network processing."
 
Previously, the device was called Fathom and its prototype was announced in April last year. Then, Intel bought this chipmaker company in September and the project went out of sight. Now, it is back with a new name with a more trustworthy seal.
 
Source: intel
 
Let’s now see the technical side of this new amazing device. There is a Myriad 2 Vision Processing Unit (VPU) in the aluminium body, that is a low-power processor consuming only 1 watt. It runs vision algorithms like facial recognition and object detection on its twelve parallel cores. The device, as per the company, runs Caffe framework based neural networks natively and delivers more than 100 gigaflops.
 
 
Source: Movidius
 
The device is extremely useful for AI researchers as it will boost the power of training and building the neural networks. Movidius explains that these sticks can be chained together to linearly accelerate the performance. For companies looking for AI power in their products, this device can be useful as it brings them an easy way to execute neural networks locally.
 
The device is available for a very nominal price of $79. Well, with such a wonderful device, artificial intelligence will be more easily accessible to the world. For more details, you can visit the Intel newsroom or the Movidius website
Up Next