EyeCloud’s OpenNCC Neural Compute Board will be a Drop-in Open-Source Replacement for Intel’s Neural Compute Stick 2 In Edge-AI Applications

EyeCloud’s OpenNCC Neural Compute Board will be a Drop-in Open-Source Replacement for Intel’s Neural Compute Stick 2 In Edge-AI Applications

Edge AI specialist, EyeCloud.AI, is getting set to launch a new OpenNCC Neural Compute Board (NCB) that will act as an open-source alternative to Intel’s Neural Compute Stick 2 (NCS2) in edge AI and computer vision projects.

The accelerated AI reference platform will support on-board inference AI models and work well with Raspberry Pi boards as well as other Arm Linux SoCs. Speaking on the design of the upcoming board and how it is expected to function, particularly on how it completely replaces the Intel Neural Compute Stick 2,  EyeCloud explained that the OpenNCC NCB has the same functionality as the Intel NCS2 with a host machine.

“The Host App on the Host Machine obtains the video stream (local file, IPC, webcam, or V4L2 MIPI cam, for example) from the outside, configures the preprocessing module according to the resolution and format of the input file of the reasoning model, then sends the visual stream to the OpenNCC NCB through the OpenNCC Native SDK for reasoning and returns the reasoning results. The reasoning supports asynchronous and synchronous models, the reasoning pipeline on OpenNCC SoM is configured through OpenNCC model JSON.”

The only difference there probably is between the OpenNCC NCB board and the Intel NCS2 is just that the former is offered as an open-source design and comes in two variants: an FPC version or a USB Type-C version which will support USB 3.0 speeds.

The compact accelerator board has at its heart, Intel’s most advanced VPU, the Movidius Myriad X VPU programmable with the Intel Distribution of the OpenVINO toolkit for porting neural networks to the edge. The VPU is the first of its kind to come with the Neural Compute Engine – a dedicated hardware accelerator for deep neural network inferences. The Neural Compute Engine works hand in hand with 16 SHAVE cores and an ultra-high throughput intelligent memory fabric to make the VPU the industry leader for on-device deep neural networks and computer vision projects. Other upgrades that have equally come with the VPU in recent times include upgraded and expanded vision accelerators, a new native 4K ISP pipeline with support for up to 8 HD sensors directly connected to the VPU.

Other board specifications will Include:

  • VPU: Intel Movidius Myriad X MV2085
  • 16 SHAVEs Vector processors
  • 20+ Vision Accelerators
  • 2x Neural Compute Engines (up to 4TOPS)
  • 8 GB, LPDDR4 (1600 MHz, 32-bit)
  • USB 3.0/USB 2.0 with USB Type-C or USB 2.0 with FPC
  • Power Supply: 5V/2A
  • Dimensions: 38 mm x 38 mm (1.5 inches by 1.5 inches)
  • Weight: 9 grams
  • Compatible with Intel’s Distribution of OpenVINO (Version 2020.3, 2021.4)

The OpenNCC NCB will also be able to support up to 6-way reasoning pipeline configuration locally and can execute a 2-way pipeline simultaneously in real-time. It can also work with multiples of itself and the SDK can dynamically allocate computing power if you wish to expand the number of NCB to suit your computing power needs.

OpenNCC NCB will be open hardware, but only hardware board schematics and open software in the form of OpenVINO and OpenNCC Frame have been published for now. The board will soon launch on CrowdSupply, so you can sign up to get notified when the campaign finally goes live as well as to get useful updates about the project.

Please follow and like us:
Pin Share
About Emmanuel Odunlade

Hardware Design Engineer | #IoT Consultant |All things #ML | Entrepreneur | Serial Writer | Passionate about Innovation and technology as tools for solving problems in developing countries. Spare time is spent around writing and advocacy for the growth of the Maker/DIY Culture in Africa.

view all posts by emmanuel
Notify of

Inline Feedbacks
View all comments
Get new posts by email:
Get new posts by email:

Join 97,426 other subscribers