TinyML Part 1: An Easy Implementation of Artificial Neural Network on an Arduino Boards
HTML-код
- Опубликовано: 2 окт 2024
- A simple RGB color classifier implementation for Arduino AVR boards and other compatible microcontrollers. This repository makes use of the Neurona library for Arduino, which enables the use of a Multilayer Perceptron (MLP) / Feed-forward Neural Network, a Machine Learning algorithm that can be run on Arduino boards. This library comes with the MLP topology workbench, which is used to train the model, and the Neurona library, which is used to convert the model into simple C++ code. Thanks to Caio Benatti Moretti's brilliant work.
Neurona library
www.arduino.cc...
MLP topology workbench
www.moretticb.c...
The code is available on Github
github.com/Tro...
#TinyML #Arduino #NeuralNetworks
Hey could anyone tell me the sensors used for the project
So is Neurona like TinyML?
And is the idea that you must collect the data (audio, images, or sensor data) via the embedded device, but the training/learning is always done on the cloud (ex. Edge Impulse) or a PC (tensorflow, keras, etc)?
And how do you tune the model for the particular application (or the type of data you are expecting)?
And what does does the output of the training/model from Neurona look like? Will it work only on the embedded device that the training data was captured on? I assume the CPU/speed will determine how fast the algorithm will work, but what are the memory requirements?
Can you offer the Artificial Neural Network code ? thank you
Good project sir. I also doing a project related with ANN. But i was using EEG sensor. So can you help me in this. And i also using arduino to do it. I'm quite poor in my coding, maybe you can help me in that. Thank you.
It depend how you make your data suitable for microcontroller, you can still use this method, make your raw data in series format in the whole window. The number of data samples per window will be your input neuron. Another method is feature's extraction, process your data using statistical analysis (e.g., max, min, stdv, etc.), this will be your new predictors, it will reduce the number of data to be feed in the network, technically saving the memory of the microcontroller. Another method is to use 32-bit MCU, you can easily develop your model using democratize AI platform, there are a lot of project out there.
TinyML is compatible with arduino uno?
Great and awesome! Could you tell us the background music? It sounds great too!
sir. it can we used another models? like svm.
Yes. you can, use different framework, called MicroML, it supports various ML algorithms using Scikit-learn
wow，it‘s cool，can you test using the other cards to test the nn model？
Code please bro
The code available here github.com/TronixLab/Easy-ArduinoNeuralNetwork
Hello dear,
Can you explain how this method use with dht11 sensor for weather detection??
It's very urgent....
dht11 sensor can only collects 3 data, temp., humidity, and dew point, if you can get more data much better, (light intensity, pressure, wind speed, etc...) to interpret the meaning of weather.
@@tronixlab7191 thank for your reply
Thx. this video is awesome !
Can you give me the code please
What i dont understand, i why you set your weights the way you did. this video is really awesome thank you!
The weights are automatically generated during the training phase, and the value may varies depend on the network architecture. You will use the generated weights (i.e., which is equivalent to the prediction model), to deploy it on your Arduino board and any support microcontroller unit. Thank you for watching.
@@tronixlab7191 Thank you! Im totally trying this.
You are using colour sensor module
Don't fool the people neral network code cannot run in Arduino
I get the RGB raw values, and use it to run with neural network with Arduino, I can give you the code if you wish.
And where he claimed he doesn't do it?
It's about processing raw data into labeled info. Ofc he didn't pick color without sensor. Wtf