This repository contains tokenized datasets generated with musicaiz library to train DL sequence models.
The data is organized as follows:
📦musicaiz-datasets
┣ 📂dataset_name
┃ ┗ 📂tokenizer
┃ ┗ 📂tokenization_type
┃ ┗ 📂train
┃ ┃ ┗ 📜token-sequences.txt
┃ ┗ 📂validation
┃ ┃ ┗ 📜token-sequences.txt
┃ ┗ 📂test
┃ ┃ ┗ 📜token-sequences.txt
┃ ┗ 📜vocabulary.txt
┃ ...
The current tree contains:
- dataset_name: jsbchorales, maestro
- tokenizer: mmm
- tokenization_type: `4_bars` and `all_bars`
The available tokenized datasets are:
Other datasets that could be included in the future:
The bps fh dataset for harmonic analysis is:
This dataset is not tokenized since is used for harmonic analysis. We trad the notes.csv of each file and convert it to a midi file that can be loaded by packages that work with MIDI files.
This project is licensed under the terms of the AGPL v3 license.
To download the data just clone this repository:
git clone git@github.com:carlosholivan musicaiz-datasets.git
cd musicaiz-datasets
To install musicaiz, clone the repository for the latest changes or simply type pip install musicaiz.
If you use any of these datasets in your work, please cite the dataset(s) you use and musicaiz software:
@misc{hernandezolivan22musicaiz,
doi = {10.48550/ARXIV.2209.07974},
url = {https://arxiv.org/abs/2209.07974},
author = {Hernandez-Olivan, Carlos and Beltran, Jose R.},
title = {musicaiz: A Python Library for Symbolic Music Generation, Analysis and Visualization},
publisher = {arXiv},
year = {2022},
}
This is a repository that hosts processed open Source datasets to train DL models. Each of these datasets have its corresponding license. It the responsability of the users to determine whether they have permission to use the dataset under the dataset's license.
If you're a dataset author or owner and you do not want your dataset to be included in this repository, please open a GitHub issue and we will remove the dataset from this repository.