asteroid 0.1.0

PyTorch-based audio source separation toolkit

Asteroid : Audio Source Separation on steroids

Asteroid is a Pytorch-based source separation and speech enhancement API that enables fast experimentation on common datasets. It comes with a source code written to support a large range of datasets, architectures, loss functions etc... and a set of recipes to reproduce some important papers.
Asteroid is intended to be a community-based project so hop on and help us !

You use asteroid or you want to?

Please, if you have found a bug, open an issue, if you solved it, open a pull request !
Same goes for new features, tell us what you want or help us building it !
Don't hesitate to join the slack and ask questions / suggest new features there as well, we won't bite !

Table of contents

• Installation
• Tutorials
• Running a recipe
• Available recipes
• Supported datasets
• Our guiding principles

Installation

In order to install Asteroid, clone the repo and install it using pip or python :

git clone https://github.com/mpariente/asteroid
cd asteroid
# Install with pip (in editable mode)
pip install -e .
# Install with python
python setup.py install

Highlights

Few notebooks showing example usage of asteroid's features.

• Permutation invariant training wrapper PITLossWrapper .
• Filterbank API

Running a recipe

Running the recipes requires additional packages in most cases, we recommend running :

# from asteroid/
pip install -r requirements.txt

Then choose the recipe you want to run and run it !

cd egs/wham/ConvTasNet
./run.sh

More information in egs/README.md.

Recipes

• ConvTasnet (Luo et al.)
• Tasnet (Luo et al.)
• Deep clustering (Hershey et al. and Isik et al.)
• Chimera ++ (Luo et al. and Wang et al.)
• FurcaNeXt (Shi et al.)
• DualPathRNN (Luo et al.)
• Two step learning (Tzinis et al.)

Supported datasets

• WSJ0-2mix / WSJ03mix (Hershey et al.)
• WHAM (Wichern et al.)
• WHAMR (Maciejewski et al.)
• Microsoft DNS Challenge (Chandan et al.)
• SMS_WSJ (Drude et al.)
• MIUSDB18 (Raffi et al.)
• FUSS (Wisdom et al.)

Writing your own recipe

Contributing

See our contributing guidelines.

Codebase structure

├── asteroid                 # Python package / Source code
│   ├── data                 # Data classes, DalatLoaders maker.
│   ├── engine               # Training classes : losses, optimizers and trainer.
│   ├── filterbanks          # Common filterbanks and related classes.
│   ├── masknn               # Separation building blocks and architectures.
│   └── utils.py
├── examples                 # Simple asteroid examples 
└── egs                      # Recipes for all datasets and systems.
│   ├── wham                 # Recipes for one dataset (WHAM) 
│   │   ├── ConvTasNet       # ConvTasnet systme on the WHAM dataset.
│   │   │   └── ...          # Recipe's structure. See egs/README.md for more info
│   │   ├── Your recipe      # More recipes on the same dataset (Including yours)
│   │   ├── ...
│   │   └── DualPathRNN
│   └── Your dataset         # More datasets (Including yours)

Remote TensorBoard visualization

# Launch tensorboard remotely (default port is 6006)
tensorboard --logdir exp/tmp/lightning_logs/ --port tf_port

# Open port-forwarding connection. Add -Nf option not to open remote. 
ssh -L local_port:localhost:tf_port user@ip

Then open http://localhost:local_port/. If both ports are the same, you can click on the tensorboard URL given on the remote, it's just more practical.

Guiding principles

• Modularity. Building blocks are thought and designed to be seamlessly plugged together. Filterbanks, encoders, maskers, decoders and losses are all common building blocks that can be combined in a flexible way to create new systems.
• Extensibility. Extending Asteroid with new features is simple. Add a new filterbank, separator architecture, dataset or even recipe very easily.
• Reproducibility. Recipes provide an easy way to reproduce results with data preparation, system design, training and evaluation in a same script. This is an essential tool for the community !