I tried that with my Aqara camera. The mic there has no windshield, though; wind noise was terrible. Also I couldn’t get Aqara support to enable higher sampling rates in the firmware than 16kHz; BirdNET expects 48kHz audio samples.
Ended up installing a better microphone attached to a RPi3A+ and using a RaspberyPi 4 which I already owned for hosting BirdNET-Go
https://maciejb.me/posts/birdnet-go-setup/
I did exactly this with BirdNet-Go and my Unifi doorbell cam. Unifi exposes an RTSP feed for each camera so it was easy for the tool to just "listen" to the doorbell and start classifying.
I have a spare e-ink display and my next weekend project to follow onto this is to wire it up so it shows some faux "woodcut" images of birds detected, or something like that.
Love your eink example! You’ve just sent me on a parts shopping trip :)
I’ve been unhappy with my little lcd board I’ve been using for the same here https://github.com/simonjgreen/OpenObservatory as the contrast is bad. I was considering using one of my trmnl displays, but it isn’t quite the right operating model for the frequency of updates I want.
The journey to a magical, free bird sound ID app involves some obvious steps (the NSF gave them money), but the amount of timeline offramps is really high. The NSF didn't think this app was educational ("it's just telling them the answer, they don't even have to work for it!"), the funding ran out so they were thinking of charging money, eBird happened to have five years of bird sound recordings they could use, a computer vision guy on the team happened to learn that spectrograms are basically the same as pictures...
And the result is tremendously useful! They get great signal on where birds are, and they're making people so much more interested in birds and nature and science. And I think charging $1 for the app or, shudder, some sort of subscription model might've ruined the whole thing.
Merlin got me into birding, and I've found this to be true, at least in some regions, especially if there's lots of background noise.
Unless a species has a really, really unique call or song, I've made it my personal rule to get eyes on it and, if I'm still not convinced, to take a picture and cross reference with iNaturalist and what folks on there think of it.
I've learned to curse the species that differ from others by the sheen of the backs of the males' necks or the angle of a wingbar. Also, gulls that only differ by things like the color of their feet or beaks.
Cornell / Merlin / eBird publishes a global species frequency DB indexed by location and date (so, how likely is a species to appear in a place, on a day of the year)... In a project that I was working on, this helped a lot when used to influence the "raw" identification based on appearance alone.
You can usually tell, though. Rare birds in the wrong place or time that come up only once are problematic, but if the bird is detected repeatedly there's a better chance. You should also be on the lookout for birds that mimic other birds (like the northern mockingbird and the gray catbird.)
I've got 33. I don't know how that stacks up against other but a large percentage of them came just from my backyard. Just got added a Baltimore Oriole that showed up this week.
I'm up to 71 thanks to the joys of taking it on holiday to other countries (less joy for my family), and occasionally going places just because I know there'll be new birds there.
Personal best for a single session is 23 birds in my back garden over the course of a couple of coffees. I have though spent a fair bit of time making the garden as attractive to different types of bird as possible - mainly through having a variety of different shaped feeders with different feed in. I can recommend the NatureSpy camera feeder too [0] - though I have had to modify it a little to make the food hole smaller otherwise magpies empty the thing in a single day just chucking all the seed on the floor. Dickheads.
Nice! I never thought about using my security cameras but ended up building a regular Birdnet-Pi with a microphone. I took it one step further to make it portable so that I can use it when we hike or travel to new places. Added a little e-ink display to the front and it allows me to see Top 3 for the day, last detected, and total number for the day.
Heads up. I noticed the screenshot of your markdown card has an ASCII block that goes below the baseline. `U+2588` (Full block) tends to do that depending on a lot of things.
I made a change to the markdown card sometime ago that uses only `U+2581` through `U+ 2587` instead to avoid that issue.
Thanks - this post reminded me to enter some sightings from yesterday into eBird (yes, another Cornell Lab birding app). A pair of rainbow bee eaters that I hadn't seen in the area before.
Came to say birdnet is fantastic but, at least in Australia, it regularly identifies calls with low probability (which are often wrong / skewed for location, but sometimes correct). Meaning you still need some wits about you. That said, it can correctly identify multiple birds in a single chorus which is cool. Just need to keep an eye out for those false positives.
I run BirdNET-Go as an app on my Home Assistant i7 M710q. I use the GPU-enabled BirdNET model and Perch V2, and it works swimmingly.
Input is three RTSP streams from various Reolink PoE cameras around our house, and it is pretty cool to be able to find and listen to the birds/frogs that come.
My impression is that the hardest part is knowing whether or not you're actually doing a good job. For vocal learners (like the ovenbird example linked above), it seems to be pretty easy using embeddings from existing models (like perch or birdnet). For non-learning species (possibly including your swifts), you can use timbre clues to differentiate many individuals, but may not be able to really nail down individual identity.
Have you found a way to identify frogs? I also have a BirdNET (the older -PI, not -Go) setup, but haven't seen a way to get anything other than birds identified.
The full Perch V2 model at https://huggingface.co/tphakala/Perch-v2 can do some frogs and cicadas and such, but it is not very good. It will certainly detect it, whether the species will be right is a whole other question.
The sound that a frog makes is 'kwak' in Dutch. And that is also the name of this bird which makes a similar sound. Onomatopoeia for the frog predator.
Nice! I've tried to build something similar with Unifi cameras (https://github.com/dewey/birdhomie) and it worked pretty well, the species detection wasn't that accurate but I'll continue on that project in the winter.
I just turned my PTZ security camera into a VAR booth for my kid. Lawn (soccer) / Driveway (Floorball) / End. Live view, step back 10s, slowmo, frame step, scrubbing.
GLM 5.3 helped to root my camera so I could get proper endpoints for PTZ and stream. Claude built the api and frontend, hosted on Home Assistant.
Inspired by AvianVisitors [1] I made a similar project [2] with an android app for bird detection using Perch, and then displays the birds detected during the last 24 hours with nice art on a samsung frame tv in "art mode".
If you dont want to repurpose an old android device, my app also supports bird detections from birdnet-go and birdweather. It should be plug and play with this project.
I noticed the linked blog mentions birds and bats. I thought bat chirps required higher sample rates than the microphones normally use. I have an Audiomoth and to set it up for bats you need to use the highest sample rate for proper ID.
I like this project. I have been using security cams around my place to keep track of animals meandering across the property. It's the best use for an outdoor solar-powered webcam if you ask me. You can get a feel for which animal to expect based on time of day or other characteristics. You also recognize those who regularly visit. I have a one-eyed raccoon that has been showing up for almost two years now. It's a female and the last time she visited the monitored fountain she had 5 rambunctious raccoonlets.
Wild that I could read this article, give DeepSeek about 4 sentences of prompts and have it running on my own homelab and unifi cameras in about 10 minutes.
The future mostly sucks but some parts are pretty cool.
Been working on a similar project for doing animal behavioral analysis on cameras setup in a barn. I got YOLO running in real-time on several streams on a RTX4070. Managing the training data and doing DINO->YOLO distillation has me writing my own data management suite that I plan on OSing soon.
Core use-case : push a message every time my dog is eating my shoes.
I have this too, I was looking to get more automated call recognition after we spotted a pheasant in my (suburban) back garden. I was going to buy a dedicated mike, but then realised that birdnet go works with any rtsp audio stream.
how confident is it in the identification? i imagine sometimes the angle is or resolution is bad and it may detect a bird but not have a lot to go off of to say what kind. does it narrow the list down or just make a guess?
It looks like they are using audio. Even with poor audio on my phone I find that it works very well. Once in a blue moon I'll get an ID I am doubtful of from BirdNet but it's rare in my experience.
but the thing with AI is that it excels in creating things we would not doubt, regardless of the truth. in particular, i find it to be susceptible to avoiding doubt at the expense of truth.
though i would be hopeful for bird audio, as differentiating their calls is something the birds themselves excel at, so it could very well be a great usecase for ai
I'm also working on a little project that is a walnut frame with a waveshare eink display and esp32 that polls for latest detection and display the picture.
I’m at a very steady state now years into running a BirdNET-GO instance. Disclaimer: I’m a contributor to the project.
I’m a pretty unhinged sorta engineer, so I actually recently created a 24/7 weather TV channel that combines real-time bird alerts, weather radar, timelapses and a sky cam feed (just a UniFi camera mounted on my home’s roof PV).
I got a little bit into birds when I met the Merlin app. And I have thought about making a bird identification station, but I'm under the impression mostly the same species of birds visit where I live. I may be wrong though.
Disclaimer: I’m also using this to see how far I can push supervised pure AI coding before it turns to mega slop, so know that before you dive in with the disparaging comments :D
(i have a few https://www.birdweather.com/ devices strewn across North America and Europe because I am time poor and preferred a turnkey solution for those gifted to, but love the DIY efforts! no affiliation!)
I'm fairly satisfied with it and would recommend it to others who aren't interested in building their own solution. I like supporting the folks who are building it, although I'm unsure what the future looks like for the product. Could enshittify, could turn into Home Assistant Foundation [1] (which I also financially support). Time will tell. That's why I think its important to have open alternatives, similar to how one of the https://www.adsbexchange.com/ operators went rouge [2], sold to private equity, and others have filled the gap with new ADS-B network solutions.
This would be more interesting with video rather than audio, no? I guess it depends where your security cameras are. Are there options for that as well?
I love AI. Thank you for such great software, anybody can do it now! Just like making videos, songs, pictures, stories, etc. What a great time to be alive! Oh, and F China for trying to sabotage our data center buildouts by leveraging useful idiots.
Ot to rain on your parade. But I made videos, songs and pictures, wrote stories back in ca. 2006 when I was 16 and read my first AI book. Without AI.
I see AI as a useful tool, but doing art with it is extremely bland compared to the real thing. I can't eben begin to describe what people who let the machine do the art miss out on.
I've done essentially the same thing before (using my existing Go2RTC and Frigate setup) but the challenge I ran into is that the microphones on my cameras just aren't very good.
In the non-fucked up timeline where Hilary won in 2016, Flock cameras is a nationwide network of automatic bird identification cameras to track bird population and migration.
I just used AI(antigravity with gemini 3.6 flash, but i'm sure many others would work) to implement this and integrate it with my security cameras after reading the blog post 26 minutes ago. I suspect the author did the same thing. I also had it steal the idea from the AvianVisitors link in this thread and it displays a generated collage of the recently detected birds.
Ended up installing a better microphone attached to a RPi3A+ and using a RaspberyPi 4 which I already owned for hosting BirdNET-Go https://maciejb.me/posts/birdnet-go-setup/
Much better sound quality now!