Chinese-trad-inst / README.md
CTIM-Gen's picture
Update README.md
59a0812 verified
metadata
license: cc-by-nc-4.0
tags:
  - music-generation
  - audio
  - traditional-chinese-music
size_categories:
  - 10G<n<100G

Chinese Traditional Instrument Dataset (CTIM-Gen Training Data)

This repository contains the training and validation datasets for CTIM-Gen, a project dedicated to controllable generation of traditional Chinese instrument performances. The dataset covers instruments including Guqin(Han Chinese), Guzheng(Han Chinese), Matouqin(Mongolian), Hulusi(Dai), Pipa(Han Chinese), Yangqin(Han Chinese), Erhu(Han Chinese).

Dataset Statistics

We curated a dataset of 12,705 high-quality audio segments, totaling approximately 106 hours. The dataset covers 7 traditional Chinese instruments, with the distribution detailed below:

Instrument Count Duration (Hours) Percentage
Guzheng (古筝) 3,264 27.20 25.7%
Pipa (琵琶) 2,831 23.59 22.3%
Guqin (古琴) 2,663 22.19 21.0%
Matouqin (马头琴) 1,401 11.68 11.0%
Erhu (二胡) 1,139 9.49 9.0%
Yangqin (扬琴) 948 7.90 7.5%
Hulusi (葫芦丝) 459 3.83 3.6%
TOTAL 12,705 105.88 100.0%

⚖️ Data Source & Usage Policy

The audio data in this dataset was primarily collected from China Music Network (www.china1901.com), with a small amount supplemented from Bilibili.

Important: This dataset is provided solely for academic research purposes in the field of AI music generation. Commercial use is strictly prohibited. By downloading or using this dataset, you agree to these terms.

Structure

  • train/: Training audio files organized by instrument.
  • valid/: Validation audio files.
  • manifest/: JSONL files containing metadata (captions, duration, etc.).