birdNET analyzer updates including Raven Pro integration (1 Viewer)

[LittleBitOfBreadNoCheese]

Couple of news items I just spotted with birdNET Analyzer

1. They have updated the model (the neural network) to v2.4 about 3 weeks ago. As discussed elsewhere, this is somewhat of a fiddle if you installed via the Windows zip file, if you installed it longhand you can change the filenames in config.py.
2. It is now integrated into the commercial version of Raven aka Raven Pro (it is NOT in the free version). The integration apparently includes both static and realtime analysis. Info is in the Raven knowledgebase in these articles

Machine Learning Detector – Running Detections

ravensoundsoftware.com

and

Machine Learning Detector – Integrating TensorFlow Models

ravensoundsoftware.com

I don't have Raven Pro but might give this a whirl.

 

[LittleBitOfBreadNoCheese]

OK I couldn't resist. The supplied model is v2.2, the email says they are writing up instructions for how to update the model to v2.4. Opened up a pre-recorded .wav file, right clicked to select Learning Detection, let it run for a few minutes and got this with the species list underneath. There is also an option to connect a mike up to it (not tried yet). They also claim to have a trainable model from birdNET that we can train on non-birds if we want.

 

[LittleBitOfBreadNoCheese]

On closer inspection the bird list I got is the same as you get in birdNET-analyzer if you don't give it lat/long (and week of year). The only thing I saw that I could configure was the species list and I was prompted with birds_europe.csv so my analysis included some birds that don't occur near me. There is a GUI option to suppress entries in the list but that would be a bit tedious as the list has 720 entries and my current analyzer list is under 200 (the analyzer tells you the current total on the command line).

The bird species lists are at
C:\Users\yellowhammer!!\Raven Pro 1.6\Models\BirdNET_GLOBAL_3K_V2.2\classes\birds_europe.csv etc

 

[LittleBitOfBreadNoCheese]

The best feature of the new integration is that you can now run your Learning Detector analysis, view the results table, highlight a particular entry and immediately play back that 3s clip. This is a big time save and for me justifies the expenditure alone.