That’s a great point! I will look into it and see if there’s a way.
edit: I’ve emailed Stefan to see what changes might be needed to ensure any annotations submitted to the dataset can be used by BirdNET.
That’s a great point! I will look into it and see if there’s a way.
edit: I’ve emailed Stefan to see what changes might be needed to ensure any annotations submitted to the dataset can be used by BirdNET.
I use the original BirdNET app. And on Android.
I downloaded this app yesterday after remembering this thread, and I absolutely love it! Thank you so much for letting us know about it!
So far, I was only able to try it out on the bush-crickets in my garden (apparently we have two types!! I didn’t know that!) and the birds in front of my bedroom window, so no verdict on how well it performs on walks yet, but right now I really really like it. I always found it annoying to juggle between Merlin and og BirdNET, but I put up with it because I wanted Merlin’s live detection AND BirdNET’s option to easily save short recordings. This new app beats both of the others in both aspects imo. The option to crop recordings before saving them is a lifesaver for my storage, and the ui for reviewing a session is so much better than Merlin’s! I can actually see all the moments where each species was supposedly detected now! It actually lets me manually disregard creaky doors that allegedly sound like owls so I don’t have a bunch of false positives in my saved recordings anymore! And it doesn’t have the server/crash issues that BirdNET has had for me for the last year or so. As a layperson, I probably won’t use the point count/survey/ARU features, but it’s cool to see that they’re there as well. Can’t wait to test it out on the frogs in my local nature reserve (once this heatwave is over)!
Just following up to say I heard back from the main contact with BirdNET, Stefan, and it seems that the way I have setup annotations in iNatSpectro is already able to be used by BirdNET and would really help provide “strong labels” to the iNaturalist audio files, which are otherwise only “weakly labelled”.
The difference being…
weak label - “there’s a call from this species somewhere in this audio file” (no annotation)strong label - “in this audio file, the target species’ call is after this time, before this time, above this frequency, and below this frequency” (a bounding box)If you use live mode, consider trying survey mode as well, which is similar to live mode but is intended to be long running. Also, in survey mode, it shows a running list of all species detected (similar to Merlin), while in live mode only the “current” detection shows up. Depending on your preference, it might work better.
Am I doing something wrong? I downloaded the BirdNET Live app from the Apple app store to my iPhone, but it only IDs birds. I was just outside next to a tree with loud cicadas, and the only IDs it offered were for a couple ravens cawing in the distance.