Creating artificial intelligence software can be a complicated task if we are unable to test it in our local system without relying on the cloud, which is an option not always available to students or programming fans.
Take advantage of artificial intelligence deep learning without cloud connection
That’s why Intel has decided to make their lives easier with the new generation of “neural computer bars,” devices that are very similar to standard USB drives (connected to computers via the USB 3.0 port), which are essentially self-contained neural networks that provide developers with a cost-effective way to create deep learning projects.
The new Intel Neural Compute Stick 2 model was presented by Intel at the AI DevCon event dedicated to the development of artificial intelligence solutions, which is currently being celebrated in Beijing. As far as we know, Intel will introduce it at a price of $99.
The price is $20 higher than its predecessor (the Intel Movidius Neural Compute Stick introduced in the summer of 2017), which is now compensated by eight times higher performance, ranging from 3-4 frames per second to 20.
Intel Neural Compute Stick 2 specifications
The Neural Compute Stick 2 is slightly larger than a USB stick and connects to a standard USB 3.0 port. The device has no fan and requires no additional power supply. It can be used with pre-trained neural networks for smart cameras, IoT devices, robotics, drones and VR hardware. Intel has already shown demonstrations of the first generation Neural Compute Stick, which detects and classifies people and objects in a real-time video stream that can be used in security applications or traffic management where low latency is important.
Neural Compute Stick 2 potential users
The Neural Compute Stick 2 is initially compatible with PCs running Linux or Raspberry Pi systems, but from Intel, they confirm that support for Windows ML will soon arrive, so Microsoft users can also enjoy the processing capacity provided by its Intel Movidius Myriad X visual processing unit. The Neural Compute Stick is initially compatible with PCs running Linux systems or with Raspberry Pi systems, but Intel confirms that support for Windows ML will soon be available, so Microsoft users can also enjoy the processing capacity that it offers.
This model will also provide optimizations for the Intel OpenVINO toolkit, which is designed to facilitate code writing for later reuse on different architectures.
With the Neural Compute Stick 2, Intel claims up to 8 times better performance than its predecessor in certain situations. The Myriad X VPU has a dedicated neural computing engine, 16 programmable arithmetic units and logic for processing dual 720p video streams.
Compatibility with Movidius Myriad X
According to Intel, the Movidius Myriad X is the world’s first SOC ( System on Chip ) that includes a dedicated neural computing engine to accelerate deep learning inferences in external devices. With its integration, the Myriad X achieves a capacity of 1 TOPS (billions of operations per second).
The Engine is an internal hardware block specifically designed to operate deep, high-speed, low-power neural networks without compromising accuracy and allowing devices to see, understand and respond to their environment in real time.