storylinez/gtzan-music-genre-dataset

GTZAN Music Genre Dataset

The GTZAN Music Genre Dataset is a collection of 1000 audio tracks each 30 seconds long. It contains 10 genres, each represented by 100 tracks. The tracks are all 22050Hz Mono 16-bit audio files in .wav format.

Overview

This dataset was created in 2002 by George Tzanetakis and Perry Cook for research in automatic music genre classification. It has become a standard benchmark dataset in the music information retrieval (MIR) community.

Dataset Structure

The dataset contains 1000 audio tracks distributed across 10 music genres:

Genre	Number of Tracks	Description
Blues	100	Traditional blues music
Classical	100	Classical music compositions
Country	100	Country music tracks
Disco	100	Disco and dance music
Hip-Hop	100	Hip-hop and rap music
Jazz	100	Jazz music tracks
Metal	100	Heavy metal music
Pop	100	Popular music tracks
Reggae	100	Reggae music
Rock	100	Rock music tracks

Audio Specifications

• Format: WAV (Waveform Audio File Format)
• Duration: 30 seconds per track
• Sample Rate: 22,050 Hz
• Channels: Mono (1 channel)
• Bit Depth: 16-bit
• Total Size: ~1.2 GB

Directory Structure

kaggle_data/
├── blues/      # 100 blues audio files
├── classical/  # 100 classical audio files  
├── country/    # 100 country audio files
├── disco/      # 100 disco audio files
├── hiphop/     # 100 hip-hop audio files
├── jazz/       # 100 jazz audio files
├── metal/      # 100 metal audio files
├── pop/        # 100 pop audio files
├── reggae/     # 100 reggae audio files
└── rock/       # 100 rock audio files

Usage

This dataset is widely used for:

• Music genre classification research
• Audio signal processing experiments
• Machine learning model training and evaluation
• Music information retrieval studies

Companion Model

A pre-trained audio genre classifier model trained on this dataset is available at: storylinez/audio-genre-classifier

Loading the Dataset

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("storylinez/gtzan-music-genre-dataset")

# Access the data
for example in dataset["train"]:
    audio_array = example["audio"]["array"]
    sample_rate = example["audio"]["sampling_rate"]
    genre = example["genre"]
    print(f"Genre: {genre}, Sample Rate: {sample_rate}Hz")

Research Applications

This dataset has been used in numerous research papers and projects including:

• Automatic music genre classification
• Feature extraction from audio signals
• Deep learning approaches to music analysis
• Comparative studies of classification algorithms

License and Attribution

This dataset is released under the MIT License. When using this dataset, please cite:

@misc{tzanetakis2002gtzan,
  title={GTZAN Genre Collection},
  author={George Tzanetakis and Perry Cook},
  year={2002},
  url={http://marsyasweb.appspot.com/download/data_sets/}
}

Important Notes

Copyright Considerations

This dataset contains copyrighted music. The original GTZAN dataset was created for research purposes. Users should be aware of copyright implications when using this dataset and should respect the intellectual property rights of the original artists and record labels.

Data Quality

Some tracks in the dataset may have quality issues or mislabeling. Researchers should consider data preprocessing and validation steps when using this dataset for machine learning applications.

Ethical Use

This dataset should be used responsibly for research and educational purposes. Commercial use may require additional licensing agreements with the respective copyright holders.

Dataset Limitations

• Limited Diversity: 100 tracks per genre may not represent the full diversity of each music genre
• Time Period: Tracks may be biased toward certain time periods in music history
• Cultural Bias: The dataset may have geographical and cultural biases in music selection
• Quality Variations: Audio quality may vary between tracks

Acknowledgments

We thank George Tzanetakis and Perry Cook for creating and sharing this valuable dataset with the research community. This dataset has enabled countless research projects and continues to be a valuable resource for music information retrieval research.