I live in the EU, where I have five separate waste streams: glass, paper, plastic/metal, RVM plastic, and general trash. They all fill at different rates, while the collection schedule is essentially nondeterministic. The only reliable signal is that they arrive at 6 a.m. sharp in trucks specifically engineered to maximize acoustic output.
My solution is similarly overengineered: a vision system combined with an olfactory system feeds into a biological neural network, which monitors the bins and determines when intervention is required.
The neural network is annoyingly power-hungry, occasionally unreliable, and has yet to support OTA firmware updates. I also have a Home Assistant as a backup for when the neural network fails. It provides a gentle nudge that perhaps the bins need taking out.
Key word is "annoyingly" One plugs into the wall and only requires maintenance(electrical bill) once a month. The efficient one permanently attached to you requires maintenance several times a day.
alexaholic · · focus · HN ↗
My solution is similarly overengineered: a vision system combined with an olfactory system feeds into a biological neural network, which monitors the bins and determines when intervention is required.
The neural network is annoyingly power-hungry, occasionally unreliable, and has yet to support OTA firmware updates. I also have a Home Assistant as a backup for when the neural network fails. It provides a gentle nudge that perhaps the bins need taking out.
Shocking, I know.
azan_ · · focus · HN ↗
I don't think it's power hungry at all! Benchmark it's energy use against SOTA, you will be surprised.
somat · · focus · HN ↗
ozyschmozy · · focus · HN ↗