Audio data preprocessing and Convolutional Neural Network training algorithm for detecting drum onsets from polyphonic music.
This project was implemented as part of my Master's Thesis in University of Jyväskylä, in the department of Mathematical Information Technology.
To get a local copy up and running follow these steps.
- Python 3.6+
- Clone the repo
git clone https://github.com/jarovaisanen/DrumOnsetDetectionCNN.git
- Install pip packages
pip install -r requirements.txt
- Install CUDA 10.1 for GPU accelerated CNN training:
- Get the ENST Drum database dataset
- Recommended to setup a virtual environment
- Modify main.py to suit your needs (Hyperparameters, Paths)
- Start data pre-processing and training the CNN by running main.py
The field is open for further development
- Fork the project
- Improve the method
Distributed under the MIT License. See LICENSE
for more information.
Jaro Väisänen - [email protected]
Project Link: https://github.com/jarovaisanen/DrumOnsetDetectionCNN
Original article
- Automatic drum transcription with convolutional neural networks Céline Jacques, Axel Roebel. Automatic drum transcription with convolutional neural networks. 21th International Conference on Digital Audio Effects, Sep 2018, Aveiro, Portugal, Sep 2018, Aveiro,Portugal. hal-02018777
Code that helped me to get started