FANNduino - Fast Artificial Neural Network for Arduino

Hi,

last year I needed a neural network for my Arduino for color recognition - just give it RGB inputs from an RGB sensor and wait for output - "red", "green", "orange" - whatever. I love FANN (Fast Artificial Neural Network) library by Steffen Nissen so I decided to use it. After a week or so I had a working Arduino C++ port of vital functions of the original C version of the library. I call it FANNduino. Although the port is not finished yet (many many many functions are still missing), I want to share it with you.

You can download FANNduino here.

It is surprisingly fast and I am working on adding more of the original functions, so stay tunned for updates. I hope you will find this library useful for your projects :-)!

First of all, thank you for your work!

I've just discovered it and tried to execute the "simple sample" on my Arduino Uno. However, once the ann is trained, when it tries to test it, it comes back to the "Setup" in an infinite loop.
Maybe a memory leak?

This is the conflictive line:
float *testOutput = ann->Run(testInputs);
Without it, everything else seems to work. But I cannot use the ann :slight_smile:

Any clue?
Thank you!

Hey! The link to your code (http://fannduino.com/) seems to be broken. Do you have the code available somewhere else? I was hoping to try it out on a Teensy!