---
title: MUSDB18
canonical_url: "https://www.modelscope.cn/datasets/OmniData/MUSDB18"
md_url: "https://www.modelscope.cn/datasets/OmniData/MUSDB18.md"
repository: OmniData/MUSDB18
last_updated: 2024-07-10
license: "[MUSDB18 Custom]"
storage_size: "4.4 GB"
domain:
  - publishDate
  - mediaTypes
  - taskTypes
  - labelTypes
  - paperUrl
  - publishUrl
  - displayName
  - publisher
tasks:
  - 2017-12-17
  - Audio
  - "Audio Source Separation"
  - Classification
  - "https://arxiv.org/pdf/1804.06267.pdf"
  - "https://sigsep.github.io/datasets/musdb.html"
  - MUSDB18
  - "Music and Audio Research Lab, New York University"
downloads: 1079
stars: 0
---

# MUSDB18

> MUSDB18 - OmniData 在 ModelScope 开源的数据集。displayName: MUSDB18 labelTypes: Classification license: MUSDB18 Custom mediaTypes: Audio paperUrl: https://arxiv.org/pdf/1804.06267.pdf publishDate: "2017-12-17" publishUrl: https://sigsep.github.io/datasets/musdb.html…

OmniData/MUSDB18 是 ModelScope 魔搭社区上的2017-12-17、Audio、Audio Source Separation数据集，涉及 publishDate、mediaTypes、taskTypes 领域，存储大小 4.4 GB，采用 [MUSDB18 Custom] 许可。

- **Repository**: OmniData/MUSDB18
- **License**: [MUSDB18 Custom]
- **Tasks**: 2017-12-17, Audio, Audio Source Separation, Classification, https://arxiv.org/pdf/1804.06267.pdf, https://sigsep.github.io/datasets/musdb.html, MUSDB18, Music and Audio Research Lab, New York University
- **Domain**: publishDate, mediaTypes, taskTypes, labelTypes, paperUrl, publishUrl, displayName, publisher
- **Storage size**: 4.4 GB
- **Downloads**: 1079
- **Stars**: 0
- **Last updated**: 2024-07-10

Source: https://www.modelscope.cn/datasets/OmniData/MUSDB18

---

displayName: MUSDB18
labelTypes:
- Classification
license:
- MUSDB18 Custom
mediaTypes:
- Audio
paperUrl: https://arxiv.org/pdf/1804.06267.pdf
publishDate: "2017-12-17"
publishUrl: https://sigsep.github.io/datasets/musdb.html
publisher:
- Media Technology Lab.USA
- University of Montpellier
- Fraunhofer Institute for Digital Media Technology
- Music and Audio Research Lab, New York University
tags:
- Drum
- Bass
- Vocals
taskTypes:
- Audio Source Separation

---
# 数据集介绍
  ## 简介
   MUDB18 是一个包含 150 个不同类型的全长音乐曲目（约 10 小时持续时间）以及它们独立的鼓、贝司、人声和其他词干的数据集。数据集分为训练集和测试集，分别包含 100 和 50 首歌曲。所有信号均为立体声并以 44.1kHz 编码。 
  ## 引文
  
```
"@article{rafii2017musdb18,
title={MUSDB18-a corpus for music separation},
author={Rafii, Zafar and Liutkus, Antoine and St{\""o}ter, Fabian-Robert and Mimilakis, Stylianos Ioannis and Bittner, Rachel},
year={2017}
}"
```

  
## Download dataset
:modelscope-code[]{type="git"}