In this repository you may find the script with which the images of the
COMReF dataset were created. The generate.py script
takes a MusicXML file as input and a target path and it generates a
measure-level dataset from it. To get the MTN annotations use the code
from the comref_converter package.
Some additional tools are provided for data analysis
- A script to count the number of tokens of each class.
- A script to clean the data for specific notation artifacts.
- A script to check for certain difficulty priors in MusicXML files.
- Python 3.9+ with
- Open-CV
- tqdm
- comref_converter (not required for the generation script)
- An installation of Inkscape
- An installation of Verovio. Check the repository for more information on how to get it up and running.
If Inkscape or Verovio cannot be found on PATH, an exception will be raised
by the probe_verovio or the probe_inkscape functions warning you.