tearitco/makeit-how-to-deploy-a-linear-regression-on-arduino-nano-33

This repository is dedicated to the first tutorial of my YouTube channel.

MakeIT-How-to-deploy-a-linear-regression-on-Arduino-Nano-33

This repository is dedicated to the first tutorial of my YouTube channel: MakeIT. This tutorial is called "How to deploy a linear regression on Arduino Nano 33 ?" available here.

You can find below the pipeline to develope a TinyML project :

Python part

The linear regression is designed using Tensorflow and Keras with only one layer of one neurone. This network has been trained on 2 numpy arrays x and y with y=2x+1. Then this model has been exported into a none quantized tflite model. After that I have used xxd tool to create a .h file containing an hexadecimal array with the the neural network's parameters.

Here is a schema of the neural network :

Arduino part

In this part I have used VS code and platformIO IDE to deploy the model onto the Arduino nano 33 BLE sense. To do it we had to import our model, and a library to deploy it. I have choosen EloquentTinyML.h. This library uses Tensorflow lite micro core and simplify the deployement. For another tutorial I will show you how to use directly tensorflow lite micro core.

Here is the result of the inferences (quite accurate) :