https://github.com/s-omranpour/DeepMusic

A python package for high level musical data manipulation and preprocessing, making data ready to be fed to a neural network.
https://github.com/s-omranpour/DeepMusic

ai-music deep-learning midi miditoolkit music python symbolic-music

Last synced: 3 months ago
JSON representation

A python package for high level musical data manipulation and preprocessing, making data ready to be fed to a neural network.

• Host: GitHub
• URL: https://github.com/s-omranpour/DeepMusic
• Owner: s-omranpour
• License: mit
• Created: 2021-05-05T21:37:47.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2022-01-19T20:34:47.000Z (almost 3 years ago)
• Last Synced: 2024-07-05T13:25:46.193Z (4 months ago)
• Topics: ai-music, deep-learning, midi, miditoolkit, music, python, symbolic-music
• Language: Python
• Homepage:
• Size: 448 KB
• Stars: 39
• Watchers: 1
• Forks: 5
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
*** A newer version of Deepmusic package with more features and better interface is available at the `multitrack` branch. However, it is still under development ***

# DeepMusic

DeepMusic is a high level python package with following features:

 - supporting different formats like MIDI, REMI, Compound Word and pianoroll. [1, 2]

 - representing musical data in a very simple but useful way for high level music theoretic manipulations.

 - preprocessing musical data in order to feed them to neural networks (chord extraction, quantization and numericalization).

 - supporting metrics used for evaluating generated sequences. [3, 4]

## Install

### With pip

`pip install deepmusic`

### From source

```

git clone https://github.com/s-omranpour/DeepMusic

cd DeepMusic

pip install .

```

## Usage

```python

from deepmusic import MusicRepr

## reading a midi file

seq = MusicRepr.from_file('test.mid')

## displaying first 10 events

print(seq[:10])

## export to remi representation

remi = seq.to_remi(ret='token')

## export to compound word representation

cp = seq.to_cp()

print(cp.shape) ## (num_events, 8)

## splitting song's bars

bars = seq.get_bars()

print(len(bars))

```

for more details please see `examples`.

## References

[1] Pop Music Transformer: Beat-based Modeling and Generation of Expressive Pop Piano Compositions, Yu-Siang Huang, Yi-Hsuan Yang

[2] Compound Word Transformer: Learning to Compose Full-Song Musicover Dynamic Directed Hypergraphs, Wen-Yi Hsiao, Jen-Yu Liu, Yin-Cheng Yeh, Yi-Hsuan Yang

[3] The Jazz Transformer on the Front Line: Exploring the Shortcomings of AI-composed Music through Quantitative Measures, Shih-Lun Wu, Yi-Hsuan Yang

[4] [https://github.com/slSeanWU/MusDr](https://github.com/slSeanWU/MusDr)