Home

Awesome

Build Status

IMPORTANT

This pacakge is now deprecated so please use the new library: https://github.com/fastaudio/fastaudio

Fast AI Audio

This is an audio module built on top of FastAI to allow you to quickly and easily build machine learning models for a wide variety of audio applications. We are an unofficial library and have no official connection to fastai except that we are huge fans and want to help make their tools more widely available and applicable to audio.

Quick Start

Google Colab Notebook

Installation

You install the library with pip but it is recommended to clone the repo if you are new to audio so that you can follow one of the notebooks.

  1. Git Clone
git clone https://github.com/fastaudio/fastai_audio
cd fastai_audio
pip install -e .
jupyter notebook
  1. Pip
pip install git+https://github.com/fastaudio/fastai_audio@0.1

Tests

Dependencies for testing are listed in the dev section of the Pipfile. Automated CI Tests are run on travis.ci via the .travis.yml file.

Setup for testing:

$ pip install pipenv
$ pipenv --three
$ pipenv shell
$ pipenv install --dev --skip-lock

Run tests:

$ pytest

Features

Audio Transform Pre-Processors

We provide a way to resample, remove silence and segment your items before generating spectrograms.

These operations are cached and depend on each other in the order below. If you change your segment settings, you won't need to resample etc.

Traintime Features

Tutorials

Known Issues

Contributors

We are looking for contributors of all skill levels. If you don't have time to contribute, please at least reach out and give us some feedback on the library by posting in the fastai audio thread or contact us via PM @baz or @madeupmasters

Citation

 @misc{coultas_blum_bracco_2019, title={fastaudio/fastai_audio}, url={https://github.com/fastaudio/fastai_audio}, journal={GitHub}, author={Coultas Blum, Harry A and Bracco, Robert}, year={2019}, month={Jul}}