Automatic Music Transcription (AMT) Tools

Implements a customizable machine learning pipeline for AMT in PyTorch. This framework abstracts various components of the AMT task, such as the dataset(s), data formatting, feature extraction, model usage, output formatting, training, evaluation, and inference. This makes for easy modification and extension through inheritance.

The framework is a work-in-progress. Its development is ongoing to meet my evolving research needs.

Installation

Standard (PyPI)

Recommended for standard/quick usage:

pip install amt-tools

Cloning Repository

Recommended for running example scripts or making experimental changes:

git clone https://github.com/cwitkowitz/amt-tools
pip install -e amt-tools

Usage

This repository can be used for many different purposes. Please see the README.md within each subpackage for more information.

Additionally, several papers are implemented under the examples/papers subdirectory in standalone scripts which utilize the framework. These examples demonstrate the versatility of the framework and serve as guides for how one might use it.