Not the most conventional application for self-hosting, but thought it might make for a fun weekend project for someone.
BirdNET-Pi is built off the back of Cornell University’s BirdNET Sound ID neural network model (https://birdnet.cornell.edu/). It can be used to classify what birds are in your vicinity by listening to their calls.
In case y’all haven’t seen it: you can run similar software in a docker container and connect it to a standard security camera stream to get audio.
This is a very cool idea! What microphone would be well suited for this purpose?
There’s a Github discussion post which is all about that topic. Worth checking it out for advice.
It largely depends on the surrounding environment in which you are going to place the device. In my case, I need to use a shotgun due to a highway nearby.
Careful with the birds!
Thanks. This definitely goes onto the pile of things I’ll build at the new house.
Loving this sweet sweet, this is what this magazine is here for, new toy.
Wife and I are into birding locally around the house. “Babe, it can listen all day and ID what it hears” might just be the tipping point to get her onboard with a couple cameras for the house.
BirdNet-Pi is awesome. Highly recommended for anyone who likes birds. The BirdNet app for phones is also nice.
Btw, BirdNet-Pi also works fine on the non-plus Raspberry Pi 3.The default install instructions explicitly prevent installation on a Pi3 or Pi3+. If you have armv7l cpu architecture, the script just exits. I banged at it for a bit, but the tensorflow lite runtime install tripped me up. Going to look into the docker project mentioned elsewhere in this thread instead.
Actually sounds really cool, why not
