Streaming Object Detection on a Raspberry Pi
Some time ago (these writeups sure take a while) my colleagues at my employer and my client and myself have played around with the Intel Movidius Neural Compute Stick (NCS).
For my not-so-specific use case: lots of data scientists, AI researchers and machine learning engineers (or whatever term currently ranks higher on Gartner’s Hype Cycle) carry around Macbooks. So imagine my disappointment when I found out that the NCS development kit (NCSDK) given by Intel only works on Ubuntu or Raspbian Stretch.
There wasn’t anyone tutorial yet that laid out all the steps to get it completely working on a Mac. So I figured, might as well aggregate everything so anyone can follow it and hopefully save some headaches. Even if you don’t know the first thing about neural networks and deep learning but are semi-handy with computers, you can get it to work!
There are two ways to go about this:
- For the full tutorial that allows Mac users to port your own trained Caffe and TensorFlow models to the native NCS format, check out my tutorial blog on Medium. Keep in mind that the Raspberry Pi only supports Caffe models, for TensorFlow you’d need to run it on an Ubuntu machine.
- For the setup that allows you to easily run someone else’s NCS format object detection models on the Raspberry Pi, check out this github project by my colleague, Wouter Poncin.
After you’re done, you can build cool/scary stuff like this:
Disclaimer: To illustrate the necessity of ethical behaviour in AI, we built a demo of what could go wrong. Hopefully, from the tongue-in-cheek tone, you realize that neither I nor my employer supports weaponization of AI.
About Sogeti Labs
SogetiLabs gathers distinguished technology leaders from around the Sogeti world. It is an initiative explaining not how IT works, but what IT means for business.