You may be under the impression that machine learning costs thousands of dollars. While that may be true in many cases, there's a lot more to machine learning than you think.For example, what if you could shower everything with a network of inexpensive machine learning-enabled sensors? The $1 TinyML Project [Jon Nordby] You can do just that. These little boards come with an STM32-like MCU, a BLE module, a Li-ion power circuit, and some nice sensor options (accelerometer, pair of microphones, light sensor).
What can we do with these sensors? [Jon] He talked a little about some of the commercial and non-commercial applications he's worked on during his ML career, starting with how accelerometers alone can perform human presence detection, sleep tracking, personal activity monitoring, or vibration pattern sensing. I did. . When it comes to audio input, there are a variety of tasks, from detecting gunshots and applause, to tracking the coffee roasting process, to detecting audio and speech. Just a few years ago, we saw machine learning being used to comfort dogs that were barking while their owners were away.
The bottom line is you should get your hands on some of these and start experimenting with ML. Training the code may require more powerful hardware, but it's much easier if you have a network of sensors waiting for commands. Additionally, since this is an open source project, you can easily add additional features needed for your specific application.
These boards are highly cost-optimized, so it's possible to order a few dozen without spending a fortune. The $1 goal is his BOM cost, especially if you choose not to include one of the more expensive sensors. You can assemble these boards yourself or have them assembled at a factory of your choice with little increase in cost. As for the software, it works with emlearn framework.
Everything from KiCad sources to Jupyter notebooks is on GitHub. When it comes to Hackaday.io, there are five worklogs that present impressive insights. Microphone work logs alone can provide insight into microphone amplification in low power conditions while keeping costs low. Want to try some image processing tasks without being too constrained by price? This is a beautiful Pi Pico ArduCam board with a camera and his TFT screen.
