Intel recently launched Neural Compute Stick 2 (NCS 2), a device that makes it easy to build smarter AI (artificial intelligence) algorithms and computer vision applications at the edge.
The NCS 2 resembles a standard USB thumb drive that can be plugged into any Linux PC or a Raspberry Pi. When a computer vision application based on artificial neural networks (ANN) is run, the complex mathematical computation is offloaded to the special chip embedded inside the stick. The chip is designed to accelerate the computation which makes the execution of ANN much faster when compared to the traditional CPU. In other words, NCS 2 acts as a micro GPU for inferencing machine learning (ML) models.
After an ML model is completely trained, it is moved into production for identifying and classifying objects, which is termed as inference. When using Intel NCS stick, developers take a fully trained Caffe2 or TensorFlow model and convert it into a format (graph) understood by the Myriad X vision processing unit embedded within the stick. Through the SDK (Software Development Kit) and toolkits provided by Intel, only the graph is loaded into the VPU chip while running the rest of the code on the local CPU.
Based on the Intel’s Movidius Myriad X vision processing unit (VPU), which is supported by the OpenVINO toolkit, the NCS 2 affordably speeds the development of deep ANN inference applications while delivering a performance boost over the earlier generation of NCS stick. The NCS 2 enables deep neural network testing, tuning, and prototyping, so developers can move from prototyping to production leveraging a range of Intel vision accelerator form factors in real-world applications.