RENNET

Deep Learning Utilities for Audio Segmentation

RENNET is the library of useful classes, functions, parsers, etc., (and may be also application backend) for Audio Segmentation using Deep Learning.

ˈrɛnɪt

curdled milk from the stomach of an unweaned calf, containing rennin and used in curdling milk for cheese.

Installation

At the moment, RENNET is not available on PyPI as a downloadable package.

Please install it from your local copy of this repository, preferably in a local Python environment, by running the following when your shell is open at the root of your copy of this repository:

pip install .

For installing extras, you can specify one or more options from the list below as, e.g. pip install -e .[analysis,dev]

• analysis :: for data-analysis and preparation tasks alongside the package.
• test :: for testing the package.
• dev :: for development on the package (use -e to install it in editable form).

Please refer to setup.py to find the list of packages used by this library, and follow their respective guidelines to install any specific versions of these libraries (e.g. to install a GPU enabled version of Tensorflow)

Documentation

TODO

Replicating Double-Talk Detection Research

To replicate the experiments done in the full thesis on Double-Talk detection in conversations using deep learning, please follow the instructions in examples/double-talk-detection/README.md. This also serves as an example workflow of using RENNET for deep learning based research in speech segmentation.

License

Apache License 2.0