OGCP-Pilot: A Physics-Aware Guitar Chord Dataset for Robust Recognition

English | 中文介绍

OGCP (Open Guitar Chord Project) Pilot Study v1.0

660 high-fidelity guitar chord samples with physical annotations for domain-robust chord recognition.

Installation

pip install ogcp

Quick Start

from ogcp import OGCPDataset, plot_fretboard
import matplotlib.pyplot as plt

# 1. 加载数据集（首次运行自动从 Hugging Face 下载）
dataset = OGCPDataset(root_dir='dataset/raw')
print(len(dataset))  # 660

# 2. 查看数据集统计
print(dataset.statistics())

# 3. 按和弦类型过滤
dm_dataset = OGCPDataset(root_dir='dataset/raw', chord_filter=['D:min'])
print(len(dm_dataset))  # 只含 Dm 样本

# 4. 获取某个和弦的所有样本
samples = dataset.get_by_chord('C:maj')
sample = samples[0]
print(sample)  # ChordSample(C:maj, open/down, fretboard=[x, 3, 2, 0, 1, 0])

# 5. 生成把位图
fig = plot_fretboard(
    sample.fretboard,
    chord_name=sample.chord_name,
    position=sample.position
)
plt.show()

# 6. 查看和弦分布
print(dataset.get_chord_distribution())

For training and inference, see ml/ml_README.md.

Data Distribution

Dataset Statistics

中文介绍

OGCP (Open Guitar Chord Project) Pilot Study v1.0

660 个高保真吉他和弦样本，包含物理标注，用于鲁棒性和弦识别研究。

安装

pip install ogcp

快速开始

from ogcp import OGCPDataset, plot_fretboard
import matplotlib.pyplot as plt

# 1. 加载数据集（首次运行自动从 Hugging Face 下载）
dataset = OGCPDataset(root_dir='dataset/raw')
print(len(dataset))  # 660

# 2. 查看数据集统计
print(dataset.statistics())

# 3. 按和弦类型过滤
dm_dataset = OGCPDataset(root_dir='dataset/raw', chord_filter=['D:min'])
print(len(dm_dataset))  # 只含 Dm 样本

# 4. 获取某个和弦的所有样本
samples = dataset.get_by_chord('C:maj')
sample = samples[0]
print(sample)  # ChordSample(C:maj, open/down, fretboard=[x, 3, 2, 0, 1, 0])

# 5. 生成把位图
fig = plot_fretboard(
    sample.fretboard,
    chord_name=sample.chord_name,
    position=sample.position
)
plt.show()

# 6. 查看和弦分布
print(dataset.get_chord_distribution())

训练与推理说明请见 ml/ml_README.md。

数据分布

数据集统计

Citation / 引用

@dataset{qiao2026ogcp,
  author = {Qin Qiao},
  title = {OGCP Pilot: A Physics-Aware Guitar Chord Dataset},
  year = {2026},
  version = {1.0},
  publisher = {Zenodo},
  doi = {10.5281/zenodo.18979053},
  url = {https://doi.org/10.5281/zenodo.18979053}
}