torchopenl3 1.0.0

Deep audio and image embeddings, based on Look, Listen, and Learn approach Pytorch

Torchopenl3

TorchopenL3 is an open-source Python library Pytorch Support for computing deep audio embeddings.

Contributors

Please refer to the Openl3 Library for keras version.

The audio and image embedding models provided here are published as part of [1], and are based on the Look, Listen and Learn approach [2]. For details about the embedding models and how they were trained, please see:

Look, Listen and Learn More: Design Choices for Deep Audio Embeddings
Jason Cramer, Ho-Hsiang Wu, Justin Salamon, and Juan Pablo Bello.
IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), pages 3852–3856, Brighton, UK, May 2019.

Comparasion

We run torchopenl3 over 100 audio files and compare with openl3 embeddings. Below is the MAE (Mean Absolute Error) Table

Content_type	Input_repr	Emd_size	MAE
Env	Linear	512	1.1522600237867664e-06
Env	Linear	6144	1.027089645617707e-06
Env	Mel128	512	1.2094695046016568e-06
Env	Mel128	6144	1.0968088741947213e-06
Env	Mel256	512	1.1641358707947802e-06
Env	Mel256	6144	1.0069775197507625e-06
Music	Linear	512	1.173499645119591e-06
Music	Linear	6144	1.048712784381678e-06
Music	Mel128	512	1.1837427564387327e-06
Music	Mel128	6144	1.0497348176841115e-06
Music	Mel256	512	1.1619711483490392e-06
Music	Mel256	6144	9.881532906774738e-07

Installation

Install via pip

pip install git+https://github.com/turian/torchopenl3.git

Install the package with all dev libraries (i.e. tensorflow openl3)

git clone https://github.com/turian/torchopenl3.git
pip3 install -e ".[dev]"

Install Docker and work within the Docker environment. Unfortunately this Docker image is quite big (about 4 GB) because

docker pull turian/torchopenl3
# Or, build the docker yourself
#docker build -t turian/torchopenl3 .

Using TorchpenL3

To help you get started with TorchopenL3 please go through the colab file.

Acknowledge

Special Thank you to Joseph Turian for his help

[1] Look, Listen and Learn More: Design Choices for Deep Audio Embeddings
Jason Cramer, Ho-Hsiang Wu, Justin Salamon, and Juan Pablo Bello.
IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), pages 3852–3856, Brighton, UK, May 2019.

[2] Look, Listen and Learn
Relja Arandjelović and Andrew Zisserman
IEEE International Conference on Computer Vision (ICCV), Venice, Italy, Oct. 2017.