عکس s-omranpour
A python package for high level musical data manipulation and preprocessing, making data ready to be fed to a neural network.Python
موضوع‌ها
۰
فورک‌ها
۱
ستاره‌ها
۲۲
تاریخ ایجاد
۱۶ اردیبهشت ۱۴۰۰
آخرین بروزرسانی
۸ ماه قبل
لایسنس
MIT License

*** 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

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