• Welcome to BirdForum, the internet's largest birding community with thousands of members from all over the world. The forums are dedicated to wild birds, birding, binoculars and equipment and all that goes with it.

    Please register for an account to take part in the discussions in the forum, post your pictures in the gallery and more.
ZEISS DTI thermal imaging cameras. For more discoveries at night, and during the day.

birdNET Analyzer model update (1 Viewer)

LittleBitOfBreadNoCheese

Well-known member
Scotland
An updated model was released a few days ago with enhanced species probability handling,


explanation here:


I've done a fair bit of work fine tuning my setup inside RavenPro so not sure if I'm going to try it.
 
By good fortune, this is the version I have downloaded to PC in the last few days. I have no previous to compare it with, but seems to work good.
I used a minimum confidence of 0.5, and species list based on Lat / Long coordinates.

The GUI version worked ok for me, otherwise there is a command line alternative, something like:

python3 analyze.py --i inputpath/ --o outputpath/ --lat 49 --lon -1 --week 0 --sensitivity 1.0 --min_conf 0.5

I'm getting into a habit of ensuring all the sound files go onto the same area in google drive, so the command will always be the same
 
Last edited:
For the non-tech minded (i.e. Me) should I be downloading and using this ID tool as a web page ?
The normal usage is bulk processing from an audio recorder. So I just record as I walk then upload to a laptop and bulk analyse the whole session. It determines where the birdsongs are and what species. The disadvantage of course is it isn't real-time - if you heard a rarity, too late to have a look at it! I convert into meta-data like this:

Score​
Label​
Count​
0.9694​
Eurasian Blue Tit​
344​
0.9972​
Long-tailed Tit​
248​
0.9956​
Eurasian Jackdaw​
224​
0.9712​
Eurasian Magpie​
118​
0.9932​
Eurasian Nuthatch​
112​
0.9912​
Coal Tit​
80​
0.9688​
European Robin​
74​
0.8346​
Dunnock​
52​
 
Anyone tried it with records from different areas than North America or Europe, to see if it fares better than the previous version? (like Merlin)
 
They have now uploaded updated install packages for v2.4 inside Raven Pro, just download the model and follow the trivial instructions and it should appear as a selection choice alongside any others (I had only v2.2 before). There also appears to be an updated Windows installer finally (downloaded but not tried it out on my other laptop yet).

If you have been customizing your species list (or 'suppressing' species as they call it) in Raven Pro this should have been updating the file birds_europe.csv, this file is substantially expanded in the new model so changes need merging.

 
Last edited:
Anyone tried it with records from different areas than North America or Europe, to see if it fares better than the previous version? (like Merlin)
I had been using it for a time in Costa Rica and it was better at picking out less common birds than Merlin but worse at common birds like those around my yard. I didn't realize the app was still receiving improvements so hadn't used it in a while. I'll give it a test and report back.
 
is it just the MData_Model file that has been updated please ?

I have a BirdNET- Pi running, can the updated MData_Model (V2) file be re-named and copied over the one on the BirdNET Pi ?

I'm assuming the positional data/chance of a type of bird has just been updated rather than any particular logic ?
 
is it just the MData_Model file that has been updated please ?

I have a BirdNET- Pi running, can the updated MData_Model (V2) file be re-named and copied over the one on the BirdNET Pi ?

I'm assuming the positional data/chance of a type of bird has just been updated rather than any particular logic ?
Not sure its quite that small a mod, there is a github tag 'species range model update'

Showing 19 changed files with 10 additions and 4 deletions.

 
Not sure its quite that small a mod, there is a github tag 'species range model update'

Showing 19 changed files with 10 additions and 4 deletions.


yes you are quite correct - it is more complex than I thought

thanks to your link though I will see if I can manually update the model on the BirdNET Pi - worse case I can roll back
 
Anyone tried it with records from different areas than North America or Europe, to see if it fares better than the previous version? (like Merlin)

Ok did a bit of testing this afternoon. On the whole, it has improved some. It did a decent job with birds that I was able to get a quality recording of. Only one of them it didn't manage to ID (and I didn't either and found out it doesn't save the unknown ones on iOS like it did when I used it on Android...lost the recording because I only analyzed it and didn't hit save.)

Species it successfully IDd:
  • Yellow-throated Toucan
  • Broad-billed Motmot (I'd like to test this vs Keel-billed which has a near identical vocalization to see if it can distinguish it)
  • Wood Thrush
  • Cocoa Woodcreeper
  • Scaly-breasted Hummingbird
  • Buff-throated Saltator
  • Smoky-brown Woodpecker
  • Rufous-tailed Jacamar
  • Scarlet-rumped Tanager
  • Orange-chinned Parakeet
  • Olive-backed Euphonia
  • Gray-headed Chachalaca
  • Common Tody-flycatcher
Species it wrongly suggested, or at least that I didn't see nor think what it heard was that species:
  • Black-crowned Antshrike (highly uncertain)
  • Yellow-throated Warbler (highly uncertain)
  • Black-capped Pygmy-Tyrant (highly uncertain)

There were also some that it heard but couldn't ID. The ones I remember it didn't manage to ID (which in part could be due to low volume as much as ID ability) were:
  • Barred Antshrike (it was making an uncommon vocalization and I had a visual ID on it to confirm).
  • Streak-headed Woodcreeper (I suspect the volume was too low for it to reliably pick it up).
  • Clay-colored Thrush (not sure why it wasn't adding suggesting this ID, common bird).
  • Great-tailed Grackle (in the background but fairly loud, not sure about this one not being picked up).
 
Just managed to get the models update working with Raven Pro, so thanks for getting the news out there, all I need to know now is how to use BirdNET-Analyzer and BirdNET-Pi ;)
I bought a one-year subscription to RavenPro so whilst that is ticking down I've stopped using the standalone birdNET-analyzer, given the prices I will probably go back to that in due course.
 
Do you have any idea what files/paths to replace to update to BirdNET-Pi? It was easy to replace the model files in Raven Pro but I've got far less confidence with the coding with BirdNET-Pi.
 
Do you have any idea what files/paths to replace to update to BirdNET-Pi? It was easy to replace the model files in Raven Pro but I've got far less confidence with the coding with BirdNET-Pi.
That was what pi users were discussing above and claimed to have just replaced this file assuming you are at V2.4 already


I don't have a Raspberry so not sure how the directories are arranged.
 

Users who are viewing this thread

Back
Top