Source Separation Models for Audio Fingerprinting

This repository contains the code described in the publication "Enhanced TV Broadcast Monitoring with Source Separation-Assisted Audio Fingerprinting". Here you can find the code of the source separation models developed for the publication. The model checkpoints are hosted in a GDrive folder (about 600MB).

To fully reproduce the work in the publication, other repositories might be needed:

• Audfprint: https://github.com/dpwe/audfprint
• Panako / Olaf: https://github.com/JorenSix/panako (at its 2.1 version release)
• NeuralFP: https://github.com/guillemcortes/neural-audio-fp (forked from https://github.com/mimbres/neural-audio-fp)

And Datasets:

• BAF dataset: https://github.com/guillemcortes/baf-dataset
• LibriSpeech: https://www.openslr.org/94/
• MillionSongDataset (MSD): http://millionsongdataset.com/
• FSD50K: https://zenodo.org/records/4060432

🎶 👂 This repository also comes with a webpage where readers can listen to some examples of audios separated by the separation models.

Installation

The authors recommend the use of virtual environments.

Requirements:

• Python 3.6+
• Create virtual environment and install requirements

git clone https://github.com/guillemcortes/ssafp.git
cd ssafp
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Usage

Training

CUDA_VISIBLE_DEVICES=0,1 python train.py --config /path/to/cfg.yml

Inference

python predict.py --input_path /path/to/input/audios/ --model /path/to/best_model.pth --out_dir /path/to/output/dir/

License

• The code in this repository is licensed under Apache 2.0

Citation

Please cite the following publication when using the dataset:

TBC

Bibtex version:

TBC

Acknowledgements

This research is part of NextCore – New generation of music monitoring technology (RTC2019-007248-7), funded by the Spanish Ministerio de Ciencia e Innovación and the Agencia Estatal de Investigación and part of resCUE – Smart system for automatic usage reporting of musical works in audiovisual productions (SAV-20221147) funded by CDTI and the Ministerio de Asuntos Económicos y Transformación Digital. Also, it has received support from Industrial Doctorates plan of the Secretaria d’universitats i Recerca, Departament d’Empresa i Coneixement de la Generalitat de Catalunya, grant agreement No. DI46-2020.