A Repetition-based Triplet Mining Approach for Music Segmentation

This repository contains a PyTorch implementation of the paper A Repetition-based Triplet Mining Approach for Music Segmentation presented at ISMIR 2023.

The overall format based on the MSAF package.

Table of Contents

  1. Usage
  2. Requirements
  3. Citing
  4. Contact

Usage

The network can be trained with:

python trainer.py --feat_id {feature type} --ds_path {path to the dataset}

The dataset format should follow:

dataset/
├── audio                   # audio files (.mp3, .wav, .aiff)
├── features                # feature files (.npy)
└── references              # references files (.jams)

To segment tracks and save deep embeddings:

python segment.py --ds_path {path to the dataset} --model_name {trained model name} --bounds {return boundaries and segment labels}

Requirements

conda env create -f environment.yml

Citing

@inproceedings{buisson2023repetition,
  title={A Repetition-based Triplet Mining Approach for Music Segmentation},
  author={Buisson, Morgan and Mcfee, Brian and Essid, Slim and Crayencour, Helene-Camille},
  booktitle={International Society for Music Information Retrieval (ISMIR)},
  year={2023}
}

Contact

morgan.buisson@telecom-paris.fr