I have a UniFi doorbell that I chose because it's self hosted and the video stays in my home. It also easily lets you get an RTSP stream of the camera feed.
Earlier this spring we put a bird feeder outside the front door and it dawned on me I could be piping the doorbell cam into BirdNET to classify the bird calls. With an RTSP stream there's no need to mount a microphone anywhere, that comes for free from the doorbell.
My wife is the bird person in the house more so than I am, but it was still really fun to set up and watch the identifications come across.
Which part of it? I assume you mean the visualisation as that’s (if I recall correctly) the only part he doesn’t break down technically in the talk. I asked him after the talk and he said he uses Vue
I was mild hoping there was a birdnet pi image with all of the magic mostly done for something like this.
like the author I'd love to set this up for my mom and mother in law as they have the phone app but it's not as exciting unless you're sitting outside.
I recently completed a MSc with a focus on bird bioacoustics, and this problem goes for the whole ecosystem/pipeline: there are very few (non-proprietary) off-the-shelf, ready to go products. Even the most popular methods that are extolled in lots of papers and by NGOs all practically need a dev/data scientist for basic implementation
Long time BirdNET fan, but I used the Merlin app for the first time yesterday and found it much more useful [0] It’ll display multiple bird species at the same time, and highlight which song belongs to which species in real-time. Recommend giving it a shot if you haven’t!
Just to clarify - the article describes BirdNET-Pi, not the mobile app Birdnet. In the mobile app we have to record and manually select a fragment to analyze, here it's a continuous monitoring where detections are visible in real time and can be replayed.
BirdNET has some advantages -- it lets you select a segment of audio and it'll give suggestions, even if it's not extremely confident in those suggestions, meaning it can sometimes do better in noisy environments. Merlin is generally more useful, but the BirdNET app has its place for sure
I make realtime algo generated ambient music based on my Birdnet-Pi server in my backyard. Give a listen at https://birdymusic.com or search birdymusic.com in your music app for some cut tracks.
I recently added an in URL meter after getting inspired by the URL based snake game on HN last week.
Just checked the dataset, it only uses the top 1000 most common north american and european birds, so if you don't live in those regions, you may not find it particularly useful
One of my future hobby projects is developing a low-cost device that can detect pigeon calls and will automatically respond with random audio of whatever local predator they fear the most. I can't stand their monotone calls in the morning.
This is so cool. I'm using BirdNET on Android for a long time and that is awesome, but running continuous monitoring on a Pi is really interesting. I saw there was also a Home Assistant integration for it.
yes! Home Assistant integration with BirdNET will take the audio streams from your cameras; no immediate need for the BirdNET-Pi. Since cameras are often outside, there is better chance to capture some interesting bird audio.
Earlier this spring we put a bird feeder outside the front door and it dawned on me I could be piping the doorbell cam into BirdNET to classify the bird calls. With an RTSP stream there's no need to mount a microphone anywhere, that comes for free from the doorbell.
My wife is the bird person in the house more so than I am, but it was still really fun to set up and watch the identifications come across.