https://github.com/robsdedude/melody-composer

Using higher Order Markov Modells and Clustering to compose melodies similar to a training data set
https://github.com/robsdedude/melody-composer

ai artificial-intelligence clustering compose-melodies computational-creativity content-generation

Last synced: 15 days ago
JSON representation

Using higher Order Markov Modells and Clustering to compose melodies similar to a training data set

• Host: GitHub
• URL: https://github.com/robsdedude/melody-composer
• Owner: robsdedude
• License: other
• Created: 2016-10-21T15:13:04.000Z (over 8 years ago)
• Default Branch: public
• Last Pushed: 2017-05-17T16:26:36.000Z (about 8 years ago)
• Last Synced: 2025-05-07T19:09:09.752Z (15 days ago)
• Topics: ai, artificial-intelligence, clustering, compose-melodies, computational-creativity, content-generation
• Language: Python
• Homepage:
• Size: 36.1 KB
• Stars: 10
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.md

Awesome Lists containing this project

README

        
This program is part of my bachelor thesis which can be found on

[arXiv](https://arxiv.org/pdf/1612.09212.pdf). It's a machine learning algorithm

that generates monophonic melodies similar to those of a phrase annotated

data-set. In- and output are midi files (i.e. machine readable music).

Setup

=====

Use virtualenv

--------------

    virtualenv venv

    source venv/bin/activate

    export LC_ALL=C

    pip install -r requirements.txt

Troubleshooting

---------------

On Ubuntu based systems it can happen, that there are header files missing. Do:

`sudo apt-get install liblapack-dev libblas-dev`

It can also happen, that you need to install a fortran compiler. Do:

`sudo install gfortran`

After fixing these things re-run `pip install -r requirements.txt` each time.

Trainings Data

-------------

Download the [MTC-FS data set](http://www.liederenbank.nl/mtc/collections.php)

and unzip it into the folder `MTC-FS-1.0`

Then run the preporcessing

    mkdir -p MTC-FS-1.0/good_midi  # create folder to put pre-processed midi files in

    source venv/bin/activate       # activate virtualenv

    python preprocess.py           # run preprocessing

    deactivate                     # deactivate virtualenv

Run the program

===============

Start the virtual environment with

`source venv/bin/activate`

To general start the program go to this directory and run

`python main.py`

More information: `python2 main.py --help`

Other things

============

To see how the clustering is working have a look at main() in contour.py