pddpauw
commented
11 months ago thanks, Andrezj!
you are right indeed... this one is not working. Had not noticed since I'm always using V2.4; will soon post a comparative field test in the original BirdNet-Pi repo confirming with some data.
I will have a look at this, but can't promise to make this one work. In that case, I will update the Readme files for this repo.
thanks again!
Install on fresh Bookworm Lite on Pi 5 completes successfully, however in the log I get:
20:14:31--- detections = analyzeAudioData(audioData, args.lat, args.lon, week, sensitivity, args.overlap) 20:14:31--- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 20:14:31--- File "/usr/local/bin/server.py", line 297, in analyzeAudioData 20:14:31--- p = predict([sig, mdata], sensitivity) 20:14:31--- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 20:14:31--- File "/usr/local/bin/server.py", line 247, in predict 20:14:31--- INTERPRETER.invoke() 20:14:31--- File "BirdNET-Pi/birdnet/lib/python3.11/site-packages/tflite_runtime/interpreter.py", line 941, in invoke 20:14:31--- self._interpreter.Invoke() 20:14:31---RuntimeError: Select TensorFlow op(s), included in the given model, is(are) not supported by this interpreter. Make sure you apply/link the Flex delegate before inference. For the Android, it can be resolved by adding "org.tensorflow:tensorflow-lite-select-tf-ops" dependency. See instructions: https://www.tensorflow.org/lite/guide/ops_selectNode number 29 (FlexRFFT) failed to prepare.
To Reproduce Steps to reproduce the behavior:
Your build Pi 5, 8GB
**
I changed the model to BirdNET_GLOBAL_6K_V2.4_Model_FP16 in tools and now it seems to work. This should be the default I guess, it is the better model anyway.