OpenNCC NCB is an accelerated-AI reference platform designed and manufactured by EyeCloud for edge-AI applications. It is a complete, open-source replacement for Intel’s Neural Compute Stick 2 (NCS2).

A Perfect Replacement for the Intel NCS2

OpenNCC Neural Compute Board (NCB) is built with Intel’s Movidius Myriad X VPU, the same chip used in the NCS. It’s Intel’s first VPU to feature the Neural Compute Engine (NCE), a dedicated hardware accelerator for deep neural-network inference. This NCE, in conjunction with 16 powerful SHAVE cores and high-throughput intelligent memory fabric, makes Movidius Myriad X ideal for on-device deep neural networks and computer-vision applications.

This NCB also provides native support for the Intel® distribution of the OpenVINO™ Toolkit, which automatically maximizes its inference performance. Developers can quickly and easily switch to different OpenVINO versions by installing different revisions of the firmware. (As of now, it supports OpenVINO 2020.3 and 2021.4.)

Compatible with the Intel® distribution of the OpenVINO™ Toolkit

Extended Edge-AI Features & GStreamer Support

You can also use OpenNCC NCB with GStreamer by replacing the local firmware with OpenNCC Frame. Used in this way, the NCB functions as an Intel NCS2 with a host machine. The Host App on the Host Machine obtains the video stream (from a local file, IPC, webcam, or V4L2 MIPI cam, for example), configures the pre-processing 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 and returns the reasoning results. It supports both synchronous and asynchronous reasoning modes, and you can configure the reasoning pipeline on the OpenNCC SoM using an OpenNCC JSON model.

This NCB supports up to a 6-way reasoning pipeline, locally, and a 2-way pipeline can run concurrently in realtime. You can realize multi-level chain or multi-model reasoning concurrency through intermediate processing via the Host App. You can also connect multiple OpenNCC NCBs together, and the SDK will dynamically allocate computing power to accommodate your needs as they fluctuate.

Suitable For Prototyping & Mass Production Alike

While Intel’s NCS was optimized for rapid prototyping, OpenNCC NCB is equally well suited to mass production. All commercial edge-AI projects face a challenge when it comes time to scale up from the development phase. Design optimizations are only part of the solution, especially for teams that lack extensive manufacturing experience. As a result, we also provide ODM/OEM services to help OpenNCC-NCB-based projects advance from prototyping to DVT, through PVT, and all the way to mass production.

Broad Compatibility With Various Machine-Learning Frameworks

OpenNCC NCB supports the TensorFlow, Caffe, Apache MXNet, and Open Neural Network Exchange (ONNX) frameworks, as well as the PaddlePaddle framework by way of an ONNX conversion. Furthermore, its compatibility with Intel’s distribution of the OpenVINO Toolkit means you can easily deploy pre-trained models from the Open Model Zoo.

Features

• Deep-learning inference at the edge
• Support for Arm-Linux SoCs and Raspberry Pi boards
• Compatible with Intel's Distribution of OpenVINO (versions 2020.3 & 2021.4 verified)
• Support for GStreamer

Technical Specifications

• VPU: Intel Movidius Myriad X MV2085
• Vector Processors: 16 SHAVEs
• Vision Accelerators: 20+
• Neural Network Capability: Two Neural Compute Engines (Up to 4 TOPS)
• RAM: 8 Gb, LPDDR4 (1600 MHz, 32-bit)
• Power Supply: 5 V / 2 A
• Data Interface: USB 3.0 / USB 2.0 with USB Type-C
• Dimensions: 1.5" x 1.5" (38 mm x 38 mm)
• Board Weight: 0.02 lbs (9 g)

Comparisons

Support & Documentation

We are happy to provide getting-started guides, open-hardware design documents, and open-source software in the form of OpenVINO and OpenNCC Frame. We have a support forum as well, and you can reach out using the appropriate link below if you have a question.