Skip to content

A repo with code generated in the paper: Leveraging Electric Guitar Tones And Effects To Improve Robustness In Guitar Tablature Transcription Modeling

License

Notifications You must be signed in to change notification settings

robust-guitar-tabs/code

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Leveraging Electric Guitar Tones and Effects to Improve Robustness in Guitar Tablature Transcription Modeling

This repository is structured into three main folders:

  • AMT-Tools: A customized version of the Cwitkowitz AMT-Tools code (Cwitkowitz AMT-Tools), including our dataset loader, for training models and saving datasets.
  • Guitar Pro Conversor: Code to convert any pre-GP5 guitar pro files to a valid format.
  • Performance Generator: Code to create token, MIDI, and jams files from Guitar Pro tabs and audio from the generated JAMS files.

Getting Started

Clone our repository and install AMT-Tools:

git clone https://github.com/robust-guitar-tabs/code
pip install -e amt-tools

This repository includes 16 Guitar Pro examples. For a full dataset, request access from DadaGP: DadaGP.

Models Training and Testing

The python files needed for training are in AMT-Tools/amt-tools/amt_tools/examples/papers/ folder.

-guitarProFx for GuitarProFx training and GuitarSet validation

-guitarSetFx for GuitarSetFx training and GuitarSet validation

-tabcnn for original GuitarSet Training and GuitarSet validation

To run EG12SET inference with any model you must run AMT-Tools\amt-tools\examples\inference\EgSet12_inference.py file, updating the desired model filepath as indicated in line 83.

Dataset Directory Structure

Each code notebook contains instructions for use. Ensure datasets are located in the datasets directory for AMT-Tools to run properly. By default, this directory is:

DEFAULT_DATASETS_DIR = os.path.join(HOME, 'Desktop', 'Datasets')

You can change this in AMT-Tools/amt-tools/amt_tools/tools/constants.py.

Folder Structure

Except for GuitarSet, which downloads by default and has a different structure, all datasets must have two folders:

  • audios
  • jams

Performance Generation

Inside the Performance Generation folder, there's a Python notebook with code to vary the GuitarSet JAMS BPM. Point it to the correct folder for use. This was utilized before audio generation for the GuitarSetFx dataset.

EGFXSET Notes

EGFXSET contains some errors. Edit the egfxset_index_1.json file and some filenames after downloading to use all 9 effects in our code.

Issues

If issues arise, feel free to reach out to the first author.

About

A repo with code generated in the paper: Leveraging Electric Guitar Tones And Effects To Improve Robustness In Guitar Tablature Transcription Modeling

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published