Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Hegel Pedroza, Wallace Abreu, Ryan Corey, Iran Roman

TL;DR

This paper addresses the robustness of guitar tablature transcription (GTT) in the face of domain shift caused by limited data. It proposes augmenting GTT training with synthetic data generated from real guitar tones processed with audio effects, rather than purely synthetic MIDI tones. TabCNN is trained on GuitarSet with two effect-based synthetic datasets (GuitarSetFX and GuitarProFX) and evaluated on a new EGSet12 test set of professional electric guitar performances, demonstrating substantial robustness gains in multi-pitch and tablature estimation. The work provides a public data/code release and highlights real-tone, effect-rich data as a practical path to bridging training and real-world GTT scenarios.

Abstract

Guitar tablature transcription (GTT) aims at automatically generating symbolic representations from real solo guitar performances. Due to its applications in education and musicology, GTT has gained traction in recent years. However, GTT robustness has been limited due to the small size of available datasets. Researchers have recently used synthetic data that simulates guitar performances using pre-recorded or computer-generated tones and can be automatically generated at large scales. The present study complements these efforts by demonstrating that GTT robustness can be improved by including synthetic training data created using recordings of real guitar tones played with different audio effects. We evaluate our approach on a new evaluation dataset with professional solo guitar performances that we composed and collected, featuring a wide array of tones, chords, and scales.

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

TL;DR

This paper addresses the robustness of guitar tablature transcription (GTT) in the face of domain shift caused by limited data. It proposes augmenting GTT training with synthetic data generated from real guitar tones processed with audio effects, rather than purely synthetic MIDI tones. TabCNN is trained on GuitarSet with two effect-based synthetic datasets (GuitarSetFX and GuitarProFX) and evaluated on a new EGSet12 test set of professional electric guitar performances, demonstrating substantial robustness gains in multi-pitch and tablature estimation. The work provides a public data/code release and highlights real-tone, effect-rich data as a practical path to bridging training and real-world GTT scenarios.

Abstract

Guitar tablature transcription (GTT) aims at automatically generating symbolic representations from real solo guitar performances. Due to its applications in education and musicology, GTT has gained traction in recent years. However, GTT robustness has been limited due to the small size of available datasets. Researchers have recently used synthetic data that simulates guitar performances using pre-recorded or computer-generated tones and can be automatically generated at large scales. The present study complements these efforts by demonstrating that GTT robustness can be improved by including synthetic training data created using recordings of real guitar tones played with different audio effects. We evaluate our approach on a new evaluation dataset with professional solo guitar performances that we composed and collected, featuring a wide array of tones, chords, and scales.
Paper Structure (13 sections, 1 figure, 2 tables)

This paper contains 13 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Datasets for training and validation
  4. Model training, and validation
  5. The EGSet12 evaluation set
  6. Metrics
  7. Results
  8. GuitarSet cross-validation
  9. Model evaluation on the new EGSet12 test domain
  10. Qualitative comparison of TabCNN models on EGSet12
  11. Discussion
  12. Conclusion
  13. Acknowledgements

Figures (1)

  • Figure 1: Each column is a two-second EGSet12 excerpt, comparing models trained using GuitarSet, with and without GuitarProFX, against ground truth. Each circled number is a tracked note on a specific guitar fret over time, the vertical lines indicate musical beats.