nes-machine-learning

Code needed for building the data and learner for predicting chiptunes using MIDI representations

Files

data/
    raw/*
        - All of the MIDI files that have been generated and collected so
          far
    all_songs/*
        - A subset of the *raw* directory that only includes songs that
          have been manually labeled as "songs", as opposed to empty
          tracks, sound effects or other background songs.
    training_songs/*
        - A subset of the *all_songs* directory, approx 2/3 of it,
          containing songs that should be used to generate and train
          the algorithm with.
    testing_songs/*
        - A subset of the *all_songs* directory, containing the approx
          1/3 of the songs that are not in the *training_songs* directory,
          that should be used for testing the accuracy of the algorithm
    training_counts.sqlite3
        - The generated sqlite3 database of transition counts (generated
          with the *training.py* file) based off the *training_songs*
          directory

Installing

install python-midi by running:

1. cd contrib/python-midi
2. python setup.py install

install MIDIUtil by running:

1. cd contrib/MIDIUtil
2. python setup.py install

You'll also want to generate some training counts. You can do so with:

1. python training.py

Related Software

• nes2midi Windows tool for converting NSF files to MIDI. The original software is in Japanese, but there is thankfully an English translation too!
• WINE Windows DLL implementation used for running the above software on OSX and Linux. OSX instructions are available on the WINE site.