Stories interesting to Practitioners of Machine Learning and A.I.
Something really cool
So if you are dabbling with IoT and Raspberry Pi’s, you will be excited about Xnor’s binary deep learning stack for embedded computing.
They have a Python-based library on Github with pre-built models using a binary neural network which we can call a ‘BNN’ for simplicity.
These models are designed to use minimal energy and cpu resources so can be run on Raspberry Pi embedded RISC devices even with solar power.
The following article is a great description of the theory
For the code-jockey, check this Github Repo out right now.
I, personally have developed Python code on a Raspberry Pi and it was a blast. Once you have created your workflow suitable to iterative development and continuous delivery (yes, even on an embedded computer this is necessary) you find making mods and updates to your stack and codebase quite rewarding. I used PyCharm which for like $70/yr has remote server support. This can be used to create code and push it directly to the embedded device from your development stack on your PC.
With this in mind and a simple remote shell to the device, you will have the Xnor detectors up and running in short order. You will of course need a camera on your Raspberry Pi which almost every IoT hacker has. I used the GoPiGo car as an experimental platform which had everything I needed to wirelessly push code and remotely control all of it via wifi.
In the tools section, I will put together a demo based on the Xnor.ai devkit and shoot some video to prove it can be done. Keep a lookout for that in the near future.