I compiled/built the TinyML book demo using the Sparkfun Edge microcontroller, which lets you load trained deep learning models onto an extremely low-powered device for edge computing. The board runs inferences, albeit slightly inaccurately. It's a great demo that runs out of the box, but there's room for improvement...which is totally part of the fun!

https://tiktok.com/@jasonsalas671/...

Errata: I typed "slightly inaccurately" to mean you can speak "yes" and "no" to the board and it'll respond with a positive case usually 1 out of every 2 or 3 times. It's still a wickedly cool thing to demo to friends.

Fair project but ew TikTok, that information security and privacy nightmare

Cool, I've been meaning to use tinyML for autonomous cars.

@PyVic The microcontroller itself, which runs TensorFlow Lite, is a neat little and inexpensive platform. The TinyML book is pure gold. Absolutely worth the read.

There are some prefab models that ship with it, too, for doing image recognition - detecting the presence of a person in a picture, and also gesture recognition. It's a fun toy to less with over a weekend. The build process is a tad involved, but I got it running: https://medium.com/@jasonsalas671/...