I have trained a model on Antarctic blue whale and Antarctic minke whale calls and noise files. The prediction and evaluation output look great and the raven annotation files match up very well with a set of audio files. However, when I run the same audio files as data source with my model setup in the Deep Learning Pamguard module the detection of the antarctic blue whale calls are quite poor, with only a few cases where the model accurately predicts the ABW call. The Antarctic minke calls are detected very well. Can you please explain why there would be any deterioration in model performance in Pamguard compared with raven annotation outputs when predicting on the same audio files?
Hello,
I have trained a model on Antarctic blue whale and Antarctic minke whale calls and noise files. The prediction and evaluation output look great and the raven annotation files match up very well with a set of audio files. However, when I run the same audio files as data source with my model setup in the Deep Learning Pamguard module the detection of the antarctic blue whale calls are quite poor, with only a few cases where the model accurately predicts the ABW call. The Antarctic minke calls are detected very well. Can you please explain why there would be any deterioration in model performance in Pamguard compared with raven annotation outputs when predicting on the same audio files?
Thanks, Lorenzo