What Movidius is doing is really fascinating. Vision processing itself is not that new (advancing rapidly, though). What's new is developing a dedicated VPU for low power devices,That is something Movidius has zoomed in on and it's been innovating. We will see what other competitions are on the horizon.
What's fascinating is its scalable approach to image processing. 12 MIPI inputs also leaves room to develop a simple one sensor engine to 4 or more and now only trick is to develop software on proprietary engine
> now only trick is to develop software on proprietary engine--znajam951
Movidius warned me that software development would be potental users most urgent concern, but claimed it has used its dozens of software engineers to create a SDT that is easy-to-use.
There are lots of high level tools out there to do all kinds of vision processing. If the performance advantages over existing solutions is significant, providing middleware (that doesn't cause too much performance degredation) could help accelerate its adoption.
I asked Movidius about the value of middleware, and this is what they replied: "Yes, we cannot agree more that providing middleware helps accelerate developers. We have learned early on that investing resources in developing useful software libraries and reference applications really does help developers immensely. For our new Myriad 2 vision processor, we are making available libraries for image processing, computer vision, and linear algebra, as well as complex kernel functions and reference application pipelines."
I am working with MYRIAD1, for research purposes in my group for more than one year. I fully support the strategic idea of having dedicated processors for acceleration of specific hot applications. I am feel confident that MYRIAD2 will be more effective due to more advanced physical technology node and more cores. The issue that should be very critical compared with the MYRIAD1, is the reliability problems, due to deep submicron technology effects. I like working with this issue. In addition, although we are academic coolaborator, the support from Movidius people is very professional.
What are the engineering and design challenges in creating successful IoT devices? These devices are usually small, resource-constrained electronics designed to sense, collect, send, and/or interpret data. Some of the devices need to be smart enough to act upon data in real time, 24/7. Specifically the guests will discuss sensors, security, and lessons from IoT deployments.