A Convolutional Neural Network which converts piano audio to a simplified MIDI format. The final model takes as input an audio file (mono- or polyphonic) and outputs a simplified MIDI output with the corresponding notes and note duration. This output can then be reconstructed into a standard MIDI file format.

The automated conversion executed by the CNN is a step toward the larger goal of Automatic Music Transcription (AMT). AMT and Music Information Retrieval have many applications in industry, including Digital Audio Workstation software development and music recommendation systems.

To get started, download the data from the Saarland Music dataset, putting the audio in one directory named "audio" and the MIDI in another directory called "midi":

mkdir audio 
wget "http://resources.mpi-inf.mpg.de/SMD/SMD_MIDI-Audio-Piano-Music.html" -e robots=off -r -l1 -nd --no-parent -A.mp3  
mkdir midi  
wget "http://resources.mpi-inf.mpg.de/SMD/SMD_MIDI-Audio-Piano-Music.html" -e robots=off -r -l1 -nd --no-parent -A.mid

In create_dataset.py, in the main function, follow the instructions to set directory_str to the filepath where you downloaded the dataset.

Run create_dataset.py.

The file models.py contains the code for the final model.

TensorFlow install instructions:
https://www.tensorflow.org/install/

Keras install instructions:
https://keras.io/#installation

collections
keras (see instructions above for dependencies)
librosa
math
matplotlib
mido
ntpath
numpy
os
pickle
random
scikit-learn
shutil
tensorflow (see instructions above for dependencies)
time

On Windows, when loading in the audio files with librosa, the following error may arise:

raise NoBackendError()
audioread.NoBackendError

If the above error is raised, try installing FFmpeg:
https://www.wikihow.com/Install-FFmpeg-on-Windows

On Linux, the equivalent fix is:
sudo apt-get install libav-tools

Copyright (c) 2018 Lillian Neff