Just a heads up for any interested, I came across this amazing project on Github and have installed it myself : BirdNet-PI
As of now, I have it running on a separate RPi 3B+ but I’m going to be migrating to an orange pi soonish. And pushing weekly info via Apprise to HA (I haven’t yet found out how to grab this notification and parse it to the HA db).
On the GitHub discussion there is talk of creating a docker version, could this mean having a possible integration/add-on in HA?! In theory, you could emulate a whole PI in docker if the system is powerful enough. And use an RTSP stream from an outdoor mic/camera.
Anyhow for those interested please leave the host a star, it’s a pretty cool project that can have a real impact.
Here’s a pic of what it picked up yesterday :
I also run this but on a Pi4 - its a great service and I really enjoy seeing what birds are around at different times of day and month. An HA integration would be very cool - notifications, visualisation etc
On the github there’s progress on getting it to work on a VM, which is a step in the right direction. Personally I think it’s suited for docker as it mainly 5 services working in unison. Maybe one day.
I got some news. Some kind fellow has ported a birdnet-pi-esque solution to docker using docker-compose. It’s very similar as it runs the service in multiple containers rather than loads of scripts as in birnetpi.
Here’s the link : BirdCAGE
The endgame being that it becomes an add-on. At the moment it’s still in its infancy.
Following on from my last post, I’ve managed to transform it into an addon. The only hitch is that it uses docker-compose, which is not a typical usage scenario for whatever reason. And thus you have to go the single-container route and allow that container to use the docker_api (level 1 addon) and run the docker-compose-up. It all starts fine but the 2nd hitch is at shutdown, I can’t modify the “S6_KILL_FINISH_MAXTIME” value from 5 seconds, it needs at least 20s to properly run the docker-compose down command.
If anyone has some experience with docker-compose and add-ons?
Maybe having the images used in docker-compose built upon the images provided by HA?
A true integration for Home Assistant would be the ‘cats meow’ (bad bird joke ) This might offer you something as a start. I wrote this rough Python program to send the ‘heard birds’ in BirdNET-Pi (or in my case a branch off that work that is running in a Proxmox VM) via MQTT to Home Assistant.
The BirdNET-Pi project and related work is a lot of fun. The code is a bit long in the tooth IMHO, but it works pretty well. I want to make a pass at simplifying the system, but that is one of many future TODO’s. For now, this code has been working pretty well and lets me see bird info in Home Assistant. You will have to poke around the code a bit to make changes to get it running in your environment, it does not currently have any ‘configuration’ files.
Yeah. I’m taking RTSP off my unifi cameras. The acquisition and processing function great. I just wish the UI could be cleaned up a bit. It sure seems like it’s downloading every audio snippet in the table every time a page loads - this makes it crawl if you’re on a phone. It’s really only usable on a desktop. That being said, it is totally neat and I love looking through it.
Are you getting accurate results with rtsp feeds? Pretty much every match is false with mine I wish it would work but I dont think the mic is very good in the cameras
I know precisely zilch about birds, but most seem very reasonable. There are a few recurring birds which are clearly wrong and I’ve blocked (frogs and dogs registering as some kind of birds) but beyond that, a lot of finches, crows, owls and hawks which I actually see daily.
It looks as if development of BirdNet-Pi has stalled. No updates for ages and no pi5 support. All is not lost however.
A version based on the programming language GO is actively being developed.
It is still being developed but the recent pre-compiled pi release works amazingly well on a pi3 compared to the BirdNet-Pi version which basically choked the pi to death. You don’t even need to use a headless install. Woks well on the desktop version.
That is with a setup cobbled together from what I had available, a SoundBlaster USB sound card and a Yamaha YPAO microphone. Tested the sound card and it has no hum and very little hiss. The microphone is meant for room calibrating Yamah AV receivers so it is omni-directional and has a wide frequency range but it has poor sensitivity (low output, may need a pre-amp). I have a better mic on order.
That’s fantastic. I have the existing one on my HA machine tied into the mic on my Unifi cameras. It works pretty well, save the UI experience is only useable on a desktop. It seems like it pre-loads ALL of the sound files on a page with the page and that makes it crawl if you try using it on mobile.
Definitely excited to see some movement here. Thanks for the update!