Multi-sensor AI filament monitor / identifier?
-
Was chatting with some buds on Discord the other day and the thought occurred:
Perhaps it could be done to use some sort of basic machine learning model to integrate data from a bunch of different sensors in order to make a more robust filament monitor that could also identify different filaments?
I was impressed by the quality of data that Thomas Sandladerer got from his nifty filament width sensor---as well as how clearly different different spools were.
That got me thinking: what one took a couple of those filament width sensors and paired them with a few other sensors---say, light transmissivity at a couple wavelengths, capacitance, and a simple color camera under defined lighting conditions---and fed them as inputs to a model. I betcha you could train it to both recognize different filament feed error modes and filament feed rates, as well as to actually recognize particular spools after using them for the first time. Might be a clever way to keep track of filament usage too!
I don't actually have much expertise in this area, but it seemed like a fun idea, so I thought I'd toss it out there.
