
Intel brings Deep Learning to the engineering masses
The device the size of a USB memory stick contains a Myriad 2 Vision Processing Unit that provides a number crunching power of more than 100 gigaflops. This high computing power enables users to develop, test and run AI applications independently from cloud resources, advertises Intel. Supporting the popular Caffe Deep Neural Network (DNN) framework, the application focus is on Neural computing and deep learning inference. Neural network projects can be ported via the Movidius compiler to run deep-learning inference on the tiny stick.
Besides of running such applications without cloud connections, developers also can use the Neural Compute Stick to perform deep learning R&D activities and prototyping on a host computer running Linux. In the future, support for the NCS will be extended to include other platforms such as the Raspberry Pi.
The NCS is available through selected channels, including distributor RS Components. Intel labels the product “democratizing Deep Learning Application Development”; a sales price however has not been published.
More information: https://developer.movidius.com/
Related articles:
Intel buys Movidius to ramp computer vision
Movidius licenses UltraSoC monitor IP
Google’s deep learning comes to Movidius
Movidius teams with Toshiba on 3-D camera system
Neurons firing at the interim stage
