Code for “Can Audio Reveal Music Performance Difficulty? Insights From the Piano Syllabus Dataset”

This repository contains the official code and pretrained models for:

Can Audio Reveal Music Performance Difficulty? Insights From the Piano Syllabus Dataset,
IEEE Transactions on Audio, Speech and Language Processing, 2025.
Pedro Ramoneda, Minhee Lee, Dasaem Jeong, Jose J. Valero-Mas, Xavier Serra.
DOI: 10.1109/TASLPRO.2025.3539018

---

🔍 Quick Inference

To estimate difficulty from an audio file:

1Run difficulty prediction using a pretrained model:

python predict_difficulty.py

You can also try the composer_model , era_model, or multirank_model in place of the basic models.

---

📂 Repository Structure

├── models/
├── get_cqt.py
├── model.py
├── predict_difficulty.py
├── zero_shot.py
├── hidden_voices.json
├── hidden_voices_features.py
├── hidden_voices_logits.py
├── make_table_basic_model.py
├── make_table_composer_model.py
├── make_table_era_model.py
├── make_table_multirank_model.py
├── show_cqts.py
├── utils.py
├── poetry.lock
├── pyproject.toml
└── README.md

---

📊 Reproduce Paper Tables

python make_table_basic_model.py  
python make_table_composer_model.py  
python make_table_era_model.py  
python make_table_multirank_model.py

---

🧪 Zero-Shot Generalization (Hidden Voices)

python hidden_voices_features.py --config hidden_voices.json  
python hidden_voices_logits.py --config hidden_voices.json --models_dir models/  
python zero_shot.py`

---

📈 Visualize CQTs

python show_cqts.py --input cqts.npy --output cqt_plot.png

📋 How to Cite

@ARTICLE{10878288,
  author={Ramoneda, Pedro and Lee, Minhee and Jeong, Dasaem and Valero-Mas, Jose J. and Serra, Xavier},
  journal={IEEE Transactions on Audio, Speech and Language Processing},
  title={Can Audio Reveal Music Performance Difficulty? Insights From the Piano Syllabus Dataset},
  year={2025},
  volume={33},
  number={},
  pages={1129-1141},
  doi={10.1109/TASLPRO.2025.3539018}
}

---

📜 License

This project is licensed under the MIT License. See LICENSE for details.