musical_python

Creating music though LSTM Recurrent Neural Networks.

Installation and setup

python -m pip install --upgrade pip
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Starting Django Server

• Compile the abc2midi executable to convert abc notation files to midi

# change to top directory (`musical_python/`)
cd abcmidi
make
mv abc2midi ../
cd ..
chmod u+rwx abc2midi

• Migrate and start Django server

python manage.py migrate
python manage.py runserver

The application can now be accessed on the localhost.

Important Note: To make sure the midi files are playable, the http://www.midijs.net/lib/midi.js file must be allowed to run on the browser.

Optional

Midi web scraping information: Pop music midi files are available in the data folder but to download midi files from other genres, use python scrap_midi/ ./scrap_freemidi_org.py <genre> <Headless State 1/0>

Note: Only for Chrome users:

For web scraping midi files with selenium, download ChromeDriver for the appropriate Chrome version here https://sites.google.com/a/chromium.org/chromedriver/downloads and save it to the scrap_midi folder.

The author not be responsible for any issues with Copyright Infrigements that might come with downloading of midi files.

Acknowledgements and References

• I claim no rights to the abcmidi package. It was created by leesavide and is available at https://github.com/leesavide/abcmidi.
• Classical music ABC notation from the ABC version of the Nottingham Music Database maintained by James Allwright http://abc.sourceforge.net/NMD/
• Midi sound samples for pop genre downloaded from https://freemidi.org
• LSTM Networks inspired from Gaurav Sharma's GitHub repository
• ABC music foundation for its documentation
• Medium: Neural Nets for Generating Music: https://medium.com/artists-and-machine-intelligence/neural-nets-for-generating-music-f46dffac21c0
• Medium: How to Generate Music using a LSTM Neural Network in Keras https://towardsdatascience.com/how-to-generate-music-using-a-lstm-neural-network-in-keras-68786834d4c5
• Website design acquired under a free open source license from http://w3layouts.com
• Headline image under CC0 1.0 Universal (CC0 1.0) Public Domain Dedication https://pxhere.com/en/photo/1079002
• Background Photo credit: Rakutaro Ogiwara. Licensed under CC0 1.0 Universal (CC0 1.0) Public Domain Dedication https://www.flickr.com/photos/arselectronica/36350783815