The term TinyML is becoming widespread, though it's still a new concept - especially given how taxing any form of Machine Learning was on hardware just a few short years ago. However, the idea behind TinyML is simple: Take data processing and inference tasks that used to require streaming data away from sensors, and perform those tasks locally on the MCU or SoC reading the sensor.
It seems almost like science fiction that this is possible, but it is. Not only that, but by using platforms like Neuton.ai or Edge Impulse, you can gather training data, train a model, and deploy it to your device without coding at all. This brings us to the fantastic prototype TinyML Water Pipeline Clog detector by Electromaker Community member CodersCafeTech.
Using a flow detector, it can determine whether the pipe is clogged completely on the edge, using an Edge Impulse model deployed to a Wio Terminal. It's a fantastic project with great documentation, and one we talked about on this week's Electromaker YouTube Show [Timed Embed]:
While this is purely a proof of concept, it uses simple tech to solve a very real problem. TinyML will play a big part in the future of smart farms, cities, and homes. The most exciting part of this is that Edge impulse makes it possible to train models without any code whatsoever.
If you are interested in projects like these, you might be interested in the Neuton.AI Webinar that took place last year, hosted by Ian: