Implement new Fish Sound detector on Orcasound data

[veirs]

A new open source convolutional neural network (CNN) has been developed by Mouy et. al. (2024) to identify fish sounds. The authors compared a random forest classifier (45 time and frequency features) with a CNN that utilizes the Ketos CNN package and found the CNN outperformed the RF classifier. These systems were trained on labeled recordings made in our local (Canadian) waters. A detailed guide to this detector is online including a tutorial on how to setup the python environment and run the CNN model..

Applying the CNN model to Orcasound archive data and/or to real-time data would be a significant enhancement to the Orcasound universe of projects.

---