free-music-archive-full

# FMA: A Dataset for Music Analysis

[Michaël Defferrard](https://deff.ch/), [Kirell Benzi](https://kirellbenzi.com/), [Pierre Vandergheynst](https://people.epfl.ch/pierre.vandergheynst), [Xavier Bresson](https://www.ntu.edu.sg/home/xbresson).

**International Society for Music Information Retrieval Conference (ISMIR), 2017.**

> We introduce the Free Music Archive (FMA), an open and easily accessible dataset suitable for evaluating several tasks in MIR, a field concerned with browsing, searching, and organizing large music collections. The community's growing interest in feature and end-to-end learning is however restrained by the limited availability of large audio datasets. The FMA aims to overcome this hurdle by providing 917 GiB and 343 days of Creative Commons-licensed audio from 106,574 tracks from 16,341 artists and 14,854 albums, arranged in a hierarchical taxonomy of 161 genres. It provides full-length and high-quality audio, pre-computed features, together with track- and user-level metadata, tags, and free-form text such as biographies. We here describe the dataset and how it was created, propose a train/validation/test split and three subsets, discuss some suitable MIR tasks, and evaluate some baselines for genre recognition. Code, data, and usage examples are available at https://github.com/mdeff/fma.

Paper: [arXiv:1612.01840](https://arxiv.org/abs/1612.01840) - [latex and reviews](https://github.com/mdeff/paper-fma-ismir2017)

Slides: [doi:10.5281/zenodo.1066119](https://doi.org/10.5281/zenodo.1066119)

Poster: [doi:10.5281/zenodo.1035847](https://doi.org/10.5281/zenodo.1035847)

# This Pack

This is the **full** dataset, comprising a total of **106,199** clips of **untrimmed length** over **16** *unbalanced* genres totaling **8,104 hours** of audio.

Packed as Parquet files, this dataset is 593 GB in size, roughly a 34% size saving over the original ZIP file.

## Repack Notes

- 173 files were unreadable by `libsndfile / libmpg123`, these were removed.
- 202 files had licenses that were unclear on whether or not they permitted redistribution, or the full license text was unavailable. These were removed.
- Many of the remaining files had mixed or inconsistent encoding. To homogenize the dataset, all audio was re-encoded using `libmpg123`.

# License

- The [FMA codebase](https://github.com/mdeff/fma) is released under [The MIT License](https://github.com/mdeff/fma/blob/master/LICENSE.txt).
- The FMA metadata is released under [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0).
- The individual files are released under various Creative Commons family licenses, with a small amount of additional licenses. **Each file has its license attached and important details of the license enumerated.** To make it easy to use for developers and trainers, a configuration is available to limit only to commercially-usable data.

Please refer to any of the following URLs for additional details.

| Class Label | License Name | URL |
| ----------- | ------------ | --- |
| 0 | CC-BY 1.0 | https://creativecommons.org/licenses/by/1.0/ |
| 1 | CC-BY 2.0 | https://creativecommons.org/licenses/by/2.0/ |
| 2 | CC-BY 2.5 | https://creativecommons.org/licenses/by/2.5/ |
| 3 | CC-BY 3.0 | https://creativecommons.org/licenses/by/3.0/ |
| 4 | CC-BY 4.0 | https://creativecommons.org/licenses/by/4.0/ |
| 5 | CC-BY-NC 2.0 | https://creativecommons.org/licenses/by-nc/2.0/ |
| 6 | CC-BY-NC 2.1 | https://creativecommons.org/licenses/by-nc/2.1/ |
| 7 | CC-BY-NC 2.5 | https://creativecommons.org/licenses/by-nc/2.5/ |
| 8 | CC-BY-NC 3.0 | https://creativecommons.org/licenses/by-nc/3.0/ |
| 9 | CC-BY-NC 4.0 | https://creativecommons.org/licenses/by-nc/4.0/ |
| 10 | CC-BY-NC-ND 2.0 | https://creativecommons.org/licenses/by-nc-nd/2.0/ |
| 11 | CC-BY-NC-ND 2.1 | https://creativecommons.org/licenses/by-nc-nd/2.1/ |
| 12 | CC-BY-NC-ND 2.5 | https://creativecommons.org/licenses/by-nc-nd/2.5/ |
| 13 | CC-BY-NC-ND 3.0 | https://creativecommons.org/licenses/by-nc-nd/3.0/ |
| 14 | CC-BY-NC-ND 4.0 | https://creativecommons.org/licenses/by-nc-nd/4.0/ |
| 15 | CC-BY-NC-SA 2.0 | https://creativecommons.org/licenses/by-nc-sa/2.0/ |
| 16 | CC-BY-NC-SA 2.1 | https://creativecommons.org/licenses/by-nc-sa/2.1/ |
| 17 | CC-BY-NC-SA 2.5 | https://creativecommons.org/licenses/by-nc-sa/2.5/ |
| 18 | CC-BY-NC-SA 3.0 | https://creativecommons.org/licenses/by-nc-sa/3.0/ |
| 19 | CC-BY-NC-SA 4.0 | https://creativecommons.org/licenses/by-nc-sa/4.0/ |
| 20 | CC-BY-ND 2.0 | https://creativecommons.org/licenses/by-nd/2.0/ |
| 21 | CC-BY-ND 2.5 | https://creativecommons.org/licenses/by-nd/2.5/ |
| 22 | CC-BY-ND 3.0 | https://creativecommons.org/licenses/by-nd/3.0/ |
| 23 | CC-BY-ND 4.0 | https://creativecommons.org/licenses/by-nd/4.0/ |
| 24 | CC-BY-SA 2.0 | https://creativecommons.org/licenses/by-sa/2.0/ |
| 25 | CC-BY-SA 2.5 | https://creativecommons.org/licenses/by-sa/2.5/ |
| 26 | CC-BY-SA 3.0 | https://creativecommons.org/licenses/by-sa/3.0/ |
| 27 | CC-BY-SA 4.0 | https://creativecommons.org/licenses/by-sa/4.0/ |
| 28 | CC-NC-Sampling+ 1.0 | https://creativecommons.org/licenses/nc-sampling+/1.0/ |
| 29 | CC-Sampling+ 1.0 | https://creativecommons.org/licenses/sampling+/1.0/ |
| 30 | FMA Sound Recording Common Law | https://freemusicarchive.org/Sound_Recording_Common_Law |
| 31 | Free Art License | https://artlibre.org/licence/lal/en |
| 32 | Free Music Philosophy (FMP) | https://irdial.com/free_and_easy.htm |

## Total Duration by License

| License | Total Duration (Percentage) |
| ------- | --------------------------- |
| CC-BY-NC-SA 3.0 | 2768.3 hours (34.16%) |
| CC-BY-NC-ND 3.0 | 2296.4 hours (28.34%) |
| CC-BY-NC-ND 4.0 | 1018.4 hours (12.57%) |
| CC-BY-NC-SA 4.0 | 533.2 hours (6.58%) |
| CC-BY 4.0 | 377.0 hours (4.65%) |
| CC-BY-NC 3.0 | 288.9 hours (3.56%) |
| CC-BY-NC 4.0 | 232.6 hours (2.87%) |
| CC-BY 3.0 | 106.9 hours (1.32%) |
| CC-BY-SA 4.0 | 99.4 hours (1.23%) |
| CC-BY-SA 3.0 | 79.7 hours (0.98%) |
| CC-BY-NC-SA 2.0 | 65.1 hours (0.80%) |
| CC-BY-NC-ND 2.0 | 56.2 hours (0.69%) |
| CC-BY-ND 3.0 | 36.8 hours (0.45%) |
| CC-BY-ND 4.0 | 25.0 hours (0.31%) |
| CC-BY-NC-ND 2.5 | 24.2 hours (0.30%) |
| FMA Sound Recording Common Law | 19.9 hours (0.25%) |
| CC-BY-NC-SA 2.5 | 18.0 hours (0.22%) |
| CC-BY-NC 2.5 | 13.3 hours (0.16%) |
| CC0 1.0 | 10.5 hours (0.13%) |
| CC-BY 1.0 | 10.4 hours (0.13%) |
| Free Music Philosophy (FMP) | 4.4 hours (0.05%) |
| Free Art License | 2.7 hours (0.03%) |
| CC-BY 2.0 | 2.5 hours (0.03%) |
| CC-BY-NC 2.1 | 2.4 hours (0.03%) |
| CC-BY-NC-SA 2.1 | 2.3 hours (0.03%) |
| CC-BY-SA 2.0 | 1.9 hours (0.02%) |
| CC-BY-NC 2.0 | 1.6 hours (0.02%) |
| CC-BY-ND 2.5 | 1.6 hours (0.02%) |
| CC-NC-Sampling+ 1.0 | 1.4 hours (0.02%) |
| CC-BY-NC-ND 2.1 | 65.0 minutes (0.01%) |
| CC-Sampling+ 1.0 | 53.9 minutes (0.01%) |
| CC-BY-SA 2.5 | 31.8 minutes (0.01%) |
| CC-BY-ND 2.0 | 29.7 minutes (0.01%) |
| CC-BY 2.5 | 11.2 minutes (0.00%) |

# Citations

```
@inproceedings{fma_dataset,
title = {{FMA}: A Dataset for Music Analysis},
author = {Defferrard, Micha\"el and Benzi, Kirell and Vandergheynst, Pierre and Bresson, Xavier},
booktitle = {18th International Society for Music Information Retrieval Conference (ISMIR)},
year = {2017},
archiveprefix = {arXiv},
eprint = {1612.01840},
url = {https://arxiv.org/abs/1612.01840},
}
```

```
@inproceedings{fma_challenge,
title = {Learning to Recognize Musical Genre from Audio},
subtitle = {Challenge Overview},
author = {Defferrard, Micha\"el and Mohanty, Sharada P. and Carroll, Sean F. and Salath\'e, Marcel},
booktitle = {The 2018 Web Conference Companion},
year = {2018},
publisher = {ACM Press},
isbn = {9781450356404},
doi = {10.1145/3184558.3192310},
archiveprefix = {arXiv},
eprint = {1803.05337},
url = {https://arxiv.org/abs/1803.05337},
}
```

# FMA：音乐分析数据集

[米夏埃尔·德费尔（Michaël Defferrard）](https://deff.ch/)、[基雷尔·本齐（Kirell Benzi）](https://kirellbenzi.com/)、[皮埃尔·范德根斯特（Pierre Vandergheynst）](https://people.epfl.ch/pierre.vandergheynst)、[泽维尔·布雷斯松（Xavier Bresson）](https://www.ntu.edu.sg/home/xbresson)。

**国际音乐信息检索会议（International Society for Music Information Retrieval Conference, ISMIR），2017年**

> 我们介绍了免费音乐档案馆（Free Music Archive, FMA）——一个开放且易于获取的数据集，适用于评估音乐信息检索（Music Information Retrieval, MIR）领域的多项任务。该领域专注于浏览、检索与组织大规模音乐藏品。当前学界对特征学习与端到端学习的兴趣与日俱增，但受限于大规模音频数据集的匮乏，这一发展受到了制约。FMA旨在解决这一难题，提供了来自16341位艺术家、14854张专辑的106574首曲目，总容量达917 GiB，总时长343天，采用包含161个类别的分层流派分类体系。数据集提供完整高音质音频、预计算特征，以及曲目与用户级元数据、标签，还有传记等自由格式文本。本文详述了该数据集的构建方案，提出了训练/验证/测试划分方案与三个子集，讨论了适用于MIR的多项任务，并针对流派识别任务评估了若干基线模型。相关代码、数据集与使用示例可访问https://github.com/mdeff/fma获取。

论文：[arXiv:1612.01840](https://arxiv.org/abs/1612.01840) - [LaTeX源码与评审意见](https://github.com/mdeff/paper-fma-ismir2017)

幻灯片：[doi:10.5281/zenodo.1066119](https://doi.org/10.5281/zenodo.1066119)

海报：[doi:10.5281/zenodo.1035847](https://doi.org/10.5281/zenodo.1035847)

# 本数据集包
这是**完整**数据集，总计包含106199段**未剪辑**的音频片段，涵盖16个**非均衡**的音乐流派，总音频时长达8104小时。

本数据集以Parquet文件格式打包，总大小为593 GB，相较于原始ZIP文件节省了约34%的存储空间。

## 重新打包说明
- 173个文件无法被`libsndfile / libmpg123`读取，已被移除。
- 202个文件的授权协议未明确是否允许再分发，或完整授权文本无法获取，已被移除。
- 剩余文件中存在多种编码格式不一致的情况。为统一数据集规范，所有音频均通过`libmpg123`进行了重编码。

# 授权协议
- [FMA代码库](https://github.com/mdeff/fma)采用[MIT许可证](https://github.com/mdeff/fma/blob/master/LICENSE.txt)发布。
- FMA元数据采用[CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/)协议发布。
- 单个音频文件采用多种知识共享系列许可协议，另有少量其他授权协议。**每个文件均附带其授权协议，并枚举了该协议的关键条款。** 为方便开发者与训练者使用，我们提供了配置选项，可仅限制使用可商用的数据。
如需更多细节，请访问以下任一链接。

| 类别标签 | 授权协议名称 | 协议链接 |
| ----------- | ------------ | --- |
| 0 | CC-BY 1.0 | https://creativecommons.org/licenses/by/1.0/ |
| 1 | CC-BY 2.0 | https://creativecommons.org/licenses/by/2.0/ |
| 2 | CC-BY 2.5 | https://creativecommons.org/licenses/by/2.5/ |
| 3 | CC-BY 3.0 | https://creativecommons.org/licenses/by/3.0/ |
| 4 | CC-BY 4.0 | https://creativecommons.org/licenses/by/4.0/ |
| 5 | CC-BY-NC 2.0 | https://creativecommons.org/licenses/by-nc/2.0/ |
| 6 | CC-BY-NC 2.1 | https://creativecommons.org/licenses/by-nc/2.1/ |
| 7 | CC-BY-NC 2.5 | https://creativecommons.org/licenses/by-nc/2.5/ |
| 8 | CC-BY-NC 3.0 | https://creativecommons.org/licenses/by-nc/3.0/ |
| 9 | CC-BY-NC 4.0 | https://creativecommons.org/licenses/by-nc/4.0/ |
| 10 | CC-BY-NC-ND 2.0 | https://creativecommons.org/licenses/by-nc-nd/2.0/ |
| 11 | CC-BY-NC-ND 2.1 | https://creativecommons.org/licenses/by-nc-nd/2.1/ |
| 12 | CC-BY-NC-ND 2.5 | https://creativecommons.org/licenses/by-nc-nd/2.5/ |
| 13 | CC-BY-NC-ND 3.0 | https://creativecommons.org/licenses/by-nc-nd/3.0/ |
| 14 | CC-BY-NC-ND 4.0 | https://creativecommons.org/licenses/by-nc-nd/4.0/ |
| 15 | CC-BY-NC-SA 2.0 | https://creativecommons.org/licenses/by-nc-sa/2.0/ |
| 16 | CC-BY-NC-SA 2.1 | https://creativecommons.org/licenses/by-nc-sa/2.1/ |
| 17 | CC-BY-NC-SA 2.5 | https://creativecommons.org/licenses/by-nc-sa/2.5/ |
| 18 | CC-BY-NC-SA 3.0 | https://creativecommons.org/licenses/by-nc-sa/3.0/ |
| 19 | CC-BY-NC-SA 4.0 | https://creativecommons.org/licenses/by-nc-sa/4.0/ |
| 20 | CC-BY-ND 2.0 | https://creativecommons.org/licenses/by-nd/2.0/ |
| 21 | CC-BY-ND 2.5 | https://creativecommons.org/licenses/by-nd/2.5/ |
| 22 | CC-BY-ND 3.0 | https://creativecommons.org/licenses/by-nd/3.0/ |
| 23 | CC-BY-ND 4.0 | https://creativecommons.org/licenses/by-nd/4.0/ |
| 24 | CC-BY-SA 2.0 | https://creativecommons.org/licenses/by-sa/2.0/ |
| 25 | CC-BY-SA 2.5 | https://creativecommons.org/licenses/by-sa/2.5/ |
| 26 | CC-BY-SA 3.0 | https://creativecommons.org/licenses/by-sa/3.0/ |
| 27 | CC-BY-SA 4.0 | https://creativecommons.org/licenses/by-sa/4.0/ |
| 28 | CC-NC-Sampling+ 1.0 | https://creativecommons.org/licenses/nc-sampling+/1.0/ |
| 29 | CC-Sampling+ 1.0 | https://creativecommons.org/licenses/sampling+/1.0/ |
| 30 | FMA Sound Recording Common Law | https://freemusicarchive.org/Sound_Recording_Common_Law |
| 31 | Free Art License | https://artlibre.org/licence/lal/en |
| 32 | Free Music Philosophy (FMP) | https://irdial.com/free_and_easy.htm |

## 各授权协议总时长
| 授权协议 | 总时长（占比） |
| ------- | --------------------------- |
| CC-BY-NC-SA 3.0 | 2768.3 小时（34.16%） |
| CC-BY-NC-ND 3.0 | 2296.4 小时（28.34%） |
| CC-BY-NC-ND 4.0 | 1018.4 小时（12.57%） |
| CC-BY-NC-SA 4.0 | 533.2 小时（6.58%） |
| CC-BY 4.0 | 377.0 小时（4.65%） |
| CC-BY-NC 3.0 | 288.9 小时（3.56%） |
| CC-BY-NC 4.0 | 232.6 小时（2.87%） |
| CC-BY 3.0 | 106.9 小时（1.32%） |
| CC-BY-SA 4.0 | 99.4 小时（1.23%） |
| CC-BY-SA 3.0 | 79.7 小时（0.98%） |
| CC-BY-NC-SA 2.0 | 65.1 小时（0.80%） |
| CC-BY-NC-ND 2.0 | 56.2 小时（0.69%） |
| CC-BY-ND 3.0 | 36.8 小时（0.45%） |
| CC-BY-ND 4.0 | 25.0 小时（0.31%） |
| CC-BY-NC-ND 2.5 | 24.2 小时（0.30%） |
| FMA Sound Recording Common Law | 19.9 小时（0.25%） |
| CC-BY-NC-SA 2.5 | 18.0 小时（0.22%） |
| CC-BY-NC 2.5 | 13.3 小时（0.16%） |
| CC0 1.0 | 10.5 小时（0.13%） |
| CC-BY 1.0 | 10.4 小时（0.13%） |
| Free Music Philosophy (FMP) | 4.4 小时（0.05%） |
| Free Art License | 2.7 小时（0.03%） |
| CC-BY 2.0 | 2.5 小时（0.03%） |
| CC-BY-NC 2.1 | 2.4 小时（0.03%） |
| CC-BY-NC-SA 2.1 | 2.3 小时（0.03%） |
| CC-BY-SA 2.0 | 1.9 小时（0.02%） |
| CC-BY-NC 2.0 | 1.6 小时（0.02%） |
| CC-BY-ND 2.5 | 1.6 小时（0.02%） |
| CC-NC-Sampling+ 1.0 | 1.4 小时（0.02%） |
| CC-BY-NC-ND 2.1 | 65.0 分钟（0.01%） |
| CC-Sampling+ 1.0 | 53.9 分钟（0.01%） |
| CC-BY-SA 2.5 | 31.8 分钟（0.01%） |
| CC-BY-ND 2.0 | 29.7 分钟（0.01%） |
| CC-BY 2.5 | 11.2 分钟（0.00%） |

# 引用格式

@inproceedings{fma_dataset,
title = {{FMA}: A Dataset for Music Analysis},
author = {Defferrard, Micha"el and Benzi, Kirell and Vandergheynst, Pierre and Bresson, Xavier},
booktitle = {18th International Society for Music Information Retrieval Conference (ISMIR)},
year = {2017},
archiveprefix = {arXiv},
eprint = {1612.01840},
url = {https://arxiv.org/abs/1612.01840},
}

@inproceedings{fma_challenge,
title = {Learning to Recognize Musical Genre from Audio},
subtitle = {Challenge Overview},
author = {Defferrard, Micha"el and Mohanty, Sharada P. and Carroll, Sean F. and Salath"e, Marcel},
booktitle = {The 2018 Web Conference Companion},
year = {2018},
publisher = {ACM Press},
isbn = {9781450356404},
doi = {10.1145/3184558.3192310},
archiveprefix = {arXiv},
eprint = {1803.05337},
url = {https://arxiv.org/abs/1803.05337},
}