<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=12389&amp;fmt=gif">

LATEST BLOG

Movidius brings Deep-Learning to a Stick

Posted by Jason Miles

Aug 9, 2017 11:01:30 PM

movidius.jpg

Movidius has developed a $79 USB device, Movidius Neural Compute Stick, which allows developers to work with deep-learning related applications.

Movidius's deep-learning applications are currently used in a number of portable settings:

1) Security cameras

2) Drones

3) Augmented reality (AR) enabled headsets

 

Movidius introduced this USB in April 2016, and back then,it was called the Fathom Neural Compute Stick.

 

One of the advantages of moving deep-learning onto a small form factor device is that it allows developers the flexibility of working from multiple locations and in an offline setting. Developers can enable prototyping, validation, and deployment through the USB device.

The Movidius Neural Compute Stick allows developers to perform AI-enabled applications 'on the edge'. This means machine learning techniques happen natively or within the device. There are three benefits from enabling an 'on the edge' performance. It allows for lower latency, higher performance, and there's the added benefit of an extra layer of security.

In today's devices, machine learning is a two-step process

1) First, machine learning allows algorithms to train on large sets of data.

2) Second, there is inference. The algorithms are used on real-world data and the results are analyzed to determine the AI's performance.

Movidius's idea has had enough traction to gain industry attention. Last year, Intel announced plans to acquire the company.

Being under Intel's umbrella allows Movidius to gain certain advantages. During the course of one year, Movidius further improved performance of its deep-learning USB stick. These improvements would have meant additional expenses, but being under Intel has allowed Movidius to cut its MSRP price from $99 to $79.


Intel has a straight-forward strategy when it comes to keeping up with cutting-edge technology. Artificial Intelligence. Augmented reality. Machine learning. Deep-neural networks. Any company that aligns with these words is potential game for Intel. Within the past two years, Intel has acquired a number of smaller companies or formed partnerships with larger companies that share similar goals.

Intel acquired mapping company, Mobileye, primarily because the product Mobileye develops will be useful for autonomous vehicles.

Intel acquired FPGA-maker, Altera.  FPGA will form the backbone in growing tech. ADAS systems. Data storage. Data security. A high-end FPGA will be found in all three applications. Intel now produces FPGA under their own name. Intel Stratix 10 is the current leading FPGA technology.

Intel has also formed partnerships with Alibaba, Baidu, and BMW.

 

EarthTron LLC is an independent distributor of electronic components. We work with both Altera/Intel and Xilinx FPGA. Contact us for a free RFQ on your FPGA. We guarantee at least 15% off franchised distribution prices.

Access Our Special FPGA Price List

Go through our blog for the latest tech news.

Sources:

Picture from Intel

TechCruch

Topics: FPGA

Subscribe for Blog Updates!