Go Down

Topic: How to design Artificial Inteligence for Arduino projects.  (Read 340 times) previous topic - next topic

tnnguyen

How to deploy machine learning on Arduino devices

Arduino single-board microcontrollers have been widely used in many Internet of thing (IoT) applications. Arduino devices are powerful, simple to use, cheap, and have great community supports. So, it is no doubt that these devices will be great choices for many embedded applications.

The question is whether machine learning applications can be deployed directly into Arduino devices? How to design those machine learning applications for Arduino devices?

People say it is impossible as Arduino devices with limited memory resources will not be suitable for machine learning applications. However, ANSCENTER announces that ANNHUB is the first machine learning software in the world that allows you to design a proper Neural Network and even advanced Bayesian Neural Networks that can be deployed directly into Arduino devices. No dependencies required.

Most importantly, ANNHUB is free to use. Let try it yourself.

You can checkout how ANNHUB generate AI code for Arduino device at ANSCENTER YouTube channel.

Artificial Intelligence Arduino


For more information, please visit ANSCENTER blog to see how we can classify three different types of IRIS followers by using AI on the Blue Pill (STM32duino) Arduino device.

How to design AI on Arduino project




zwieblum

Nice project. But please correct your statements: It's not a "prediction"  that the arduino makes, but a "classification". A "prediction" would mean that it could predict on basis of the alredy presented datasets the next dataset you would present. But maybe I am wrong any your "AI" does not do classification on base of the current dataset, but "prediction" of the next (now unknown) dataset. This would be quite a market, all anlysts could put their anique glassspheres to a rest and simply use arduino ...

jremington

Totally misleading. What the Arduino does is basically a table lookup, the simplest and easiest part of the entire project.

All the real work is done on someone else's computer.

taterking

the first concepts for the algorithms for machine learning have been around since the 80's. the reason why they haven't been in place until somewhat recently is because of the lack of proccessing speeds and input data to make them usefull untill now.  learning algorythms are challenging for a modern desktop proccesser so with an arduino chip that is thousands of times slower you definatly wont be creating "the terminator" with an arduino processor ! at best i couldnt imagine an arduino doing much more that recognizing low def picture of the alphebet or possible playing a game of tick-tack-toe. I think an arduino processor is the wrong tool for the job. i would consider something faster like a pie or mini PC or something.

 
Building robots to take over the world.

Go Up