Paving the way for deep-learning based sound detectors and classifiers
Author: Fabio Frazao and Oliver Kirsebom - MERIDIAN Data Analytics
MERIDIAN's data analytics team is excited to announce the first release of a software package for acoustic data analysis with neural networks. The package, named Ketos after the Ancient Greek word for whale, has been developed with a particular eye to detection and classification tasks in underwater acoustics.
Ketos is written in Python and utilizes a number of powerful open-source software libraries including NumPy, HDF5, and Tensorflow. It is licensed under the GNU GPLv3 license and hence freely available for anyone to use and modify.
In essence, Ketos is a collection of Python classes and functions that make it easier to implement neural networks and train them to detect and classify sounds. This includes functions to handle and manipulate acoustic data, for example, to compute spectrograms, keep track of annotations, and store large amounts of data in easily queryable database formats.
The first version of Ketos is aimed at researchers and data scientists in underwater acoustics, who already have some knowledge of the programming language Python and a basic understanding of machine learning and neural networks. Future versions of Ketos will seek to lower the "admission requirements" by expanding the catalogue of tutorials and implementing more high-level functionalities.
If you are curious to learn more about the project, please explore the Ketos documentation page where you will find detailed descriptions of all functionalities, short code snippets, comprehensive tutorials, and installation instructions.
Also, if you wish to get involved (and we hope you do!) please consult the How to Contribute section of the documentation page, which lists all the ways in which you can contribute to the continued development of Ketos.