Yeah, I came across that. It's helpful for testing and trying things out. But, it isn't exactly something I can sell the higher ups on when making a decision on whether to integrate a device like the Neural Compute Stick into our low powered Intel Atom/Windows IoT based instrument.
For that I only see two viable options.
Option A - If Intel ever has official Windows 10 support in the NCSDK.
Option B - If Microsoft includes support for it within their ML framework stuff.
My question was whether the Windows AI platform will support the Neural Compute stick by default. Is that something Intel is actively working with Microsoft on? Maybe this a question best left for Microsoft if one could actually talk to Microsoft.
In the Verge article it says "AI processors like Intel's Movidius VPU will also be supported" so I'm assuming that includes the Neural Compute stick. If not then maybe I should consider integrating the Movidius VPU.
I did try the ML framework since it's available in the latest Windows 10 Insiders preview build. But, I didn't see anything that indicated whether the Neural Compute stick was recognized. One of the options allowed me to select between the CPU, and GPU.
Here is what I tried out.
https://github.com/Microsoft/Windows-Machine-Learning
