Introducing EG-IPT and ipt~: a novel electric guitar dataset and a new Max/MSP object for real-time classification of instrumental playing techniques.

This repository contains the code for the paper of the same name, which introduces the Electric Guitar Instrumental Playing Techniques (EG-IPT) dataset and the ipt~ Max/MSP external object.

This work was presented at the NIME 2025 conference in Canberra, Australia.

📁 Project Structure

nime2025/
├── data/
│   ├── raw/                 # Raw audio recordings
│   └── dataset/             # Store dataset csv files
├── augments.py              # Data augmentation definitions
├── externals/               # Batch sampler implementation
├── model.py                 # Model definitions
├── results/                 # Folder for NIME2025 .ts model and training results
├── utils.py                 # Dataset management, audio processing, and training functions
├── nime2025.ipynb           # Main notebook
└── requirements.txt         # Project dependencies

🚀 Getting Started

1. Download and prepare the dataset

Download the EG-IPT dataset from here.
Once downloaded, extract it and place the contents data/raw/ directory of this repository.

Make sure path point to correct location of the dataset. The directory structure should look like this:

data/raw/EG-IPT/
├── HB-neck/
├── HB-bridge/
└── HB-couple/

Paths can be modified in the notebook if needed.

2. Python environment

Make sure you have Python 3.11.11 installed. We recommend using a dedicated conda environment:

conda create --name nime2025 python=3.11.11
conda activate nime2025

3. Run the notebook

Open nime2025.ipynb in Jupyter and run all cells sequentially. It will guide you through:

• Installing required Python packages
• Generating dataset
• Performing preprocessing and augmentation
• Training the model and evaluating it
• Exporting a TorchScript .ts model for real-time usage

🎛️ Real-time Usage in Max

For real-time use in Max, check our other repository 👉 ipt_tilde

This repository provides the code necessary to compile a Max external object to run .ts models exported via our the jupyter notebook in Max.

🧠 About

This project is part of an ongoing research effort into the real-time recognition of instrumental playing techniques for interactive music systems. If you use this work in your paper, please consider citing the following:

@inproceedings{fiorini2025egipt,
  title={Introducing EG-IPT and ipt~: a novel electric guitar dataset and a new Max/MSP object for real-time classification of instrumental playing techniques},
  author={Fiorini, Marco and Brochec, Nicolas and Borg, Joakim and Pasini, Riccardo},
  booktitle={NIME 2025},
  year={2025},
  address={Canberra, Australia}
}

📚 Related Work

If you are interested in this topic, please check out our other papers:

• Fiorini and Brochec (2024) - "Guiding Co-Creative Musical Agents through Real-Time Flute Instrumental Playing Technique Recognition"
• Brochec et al. (2024) - "Microphone-based Data Augmentation for Automatic Recognition of Instrumental Playing Techniques"
• Brochec and Tanaka (2023) - "Toward Real-Time Recognition of Instrumental Playing Techniques for Mixed Music: A Preliminary Analysis"

📜 License and Fundings

This project is released under a GPL-3.0 license.

This research is supported by the European Research Council (ERC) as part of the Raising Co-creativity in Cyber-Human Musicianship (REACH) Project directed by Gérard Assayag, under the European Union's Horizon 2020 research and innovation program (GA #883313). Funding support for this work was provided by a Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) scholarship to Nicolas Brochec.