Q: What’s under 3 inches long … plugs like a thumb drive into the USB port of a Windows 10, Ubuntu/CentOS Linux or MacOS computer … includes the OpenVINO toolkit … makes it easy to develop computer-vision and AI prototypes for IoT and edge devices … yet retails for just $70?
A: Intel Neural Compute Stick 2.
Intel Neural Compute Stick 2 (NCS 2) is a development and prototyping platform. You can use it to quickly bring computer vision and AI to IoT and edge devices.
Intel NCS 2 is built on the latest Intel Movidius Myriad X VPU (visual processing unit), a dedicated hardware accelerator for deep neural-network inferences. This component, combined with access to the Intel Distribution of the OpenVINO toolkit, boosts the NCS 2’s performance by up to 8x over the previous generation.
You can combine the hardware-optimized performance of the Intel Movidius Myriad X VPU and the Intel Distribution of OpenVINO Toolkit to accelerate deep neural network-based applications to systems with a common USB Type-A port. You can also use the NCS 2 to test, tune and optimize your applications, then move from prototype to production with help from an Intel AI: In Production partner.
Intel NCS 2 supports common frameworks and includes out-of-the box sample applications for easy on-boarding and fast development. And you can run NCS 2 “at the edge” without a cloud-computing connection.
As mentioned above, Intel NCS 2 comes with a free downloadable software developer kit, the Intel Distribution of OpenVINO toolkit. It’s a collection of developer tools which can be used to develop and deploy computer vision inference applications across Intel NCS 2 and other Intel hardware.
The Intel Distribution of OpenVINO toolkit also supports frameworks such as Caffe and TensorFlow on Intel NCS 2. Also included are APIs for C++ and Python that you can use to build deep-learning applications.
Models, references included
Pretrained models are included, too, so you don’t have to search for or train your own. These include a smart shopping cart, 3D printing error detection, machine learning for mammography, and coral reef restoration and research.
There’s also a long list of open-source reference implementations to help you quickly deploy projects. These include intruder detection, store traffic and shopper gaze monitors, a parking lot tracker, and a monitor for machine operators.
And once you’ve developed your prototype on Intel NCS 2, you can bring it to production with help from the Intel AI: In Production website. It’s your ecosystem of equipment providers, software and analytics providers, systems integrations, solution aggregators and cloud service providers that can help you get rolling.