Here you can find the code for training and evaluating WhaleNet architecture.
WhaleNet architecture is a novel deep convolutional architecture designed for marine mammals vocalization, that combines Wavelet Scattering Transform and Mel Spectrogram for enhanced feature exctraction.
Python script whalenet.py contains the full pipeline, including data downloading, data preprocissing, architecture training and evaluation.
Data is public and can be downloaded from the Watkins' Marine Mammals Sound Dataset- official webpage (https://whoicf2.whoi.edu/science/B/whalesounds/index.cfm) The scripts are designed to receive the data as a zip file.
corresponding author: Alessandro Licciardi (Politecnico di Torino, Istituto Nazionale di Fisica Nucleare) alessandro.licciardi@polito.it
co-author: Davide Carbone (Politecnico di Torino, Istituto Nazionale di Fisica Nucleare) davide.carbone@polito.it