Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


MORP: Movidius On Raspberry Pi This repo about using Movidius neural compute sticks(NCS) on the Raspberry Pi.

It is the ongoing result of a series Meetups I have been hosting for Chicago Area Embedded Deep Learning:

As of 12/19/17 Movidius is saying you can have the NCS or TensorFlow, but not both:

Slides from a very short demo I gave:

FAQ: What are the Movidius links?:

How do I use the Pi Camera:

What hardwhere do I need to get into this?

  • Really you need a Ubuntu 16.04 PC, RPi, and NCS.

Does it have to be Ubuntu 16.04 PC?

  • Only if you value your time! Preferably it should be a modern PC with a GPU running Ubuntu directly(not virtual). This has been a major stumpling block both in the meetup group and on the movidous forum.

What does it do?

  • Infers(runs the computational graph) a Tensorflow or caffe model.

What are some NCS+RPI pros and cons?

  • Pros:
  • Power usage(2 Watts NCS+RPI)
  • Size and wieght
  • It's the first ASIC out so there is nothing to compare it to at this point
  • It's true edge device (no internet needed & appliance like)
  • RPI is very popular
  • 15x quicker than RPI along on Deep Learning inferance.
  • Cons:
  • First one out: so limited support & could have much more power
  • You can't access the Myraid 2 VPU with all the computer vision functionality
  • Cost is high for what your getting (It's less than one thread on a CPU)
  • If power and size are not an issue an old PC is more cost effective and easier to use.

How does the NCS compare to the Google AIY Vision Kit?

  • Not sure yet, I ordered three of them
  • The Google AIY Vision kit is a hat/bonnet that sits on top of a RPi Zero
  • The Vision Kit is a hobby/learning product only, becuase you can't buy RPI Zeros in quanity and need to solder the headers on(correctly) to use it.

**More FAQ's to Come

Thanks! Paul Krush 630-830-6640






No releases published


No packages published


You can’t perform that action at this time.