towards automatic transcription of polyphonic electric guitar music:a new dataset and a multi-loss transformer model

TL;DR

This paper introduces a new electric guitar transcription dataset, EGDB, and evaluates existing piano transcription models alongside a novel multi-loss Transformer, highlighting challenges and potential improvements in polyphonic guitar transcription.

Contribution

The paper presents EGDB, a new dataset for electric guitar transcription, and proposes a multi-loss Transformer model, benchmarking its performance against existing models.

Findings

Timbre significantly affects transcription accuracy.

Multi-loss Transformers show promise for guitar transcription.

Room for further improvement in polyphonic guitar transcription.

Abstract

In this paper, we propose a new dataset named EGDB, that con-tains transcriptions of the electric guitar performance of 240 tab-latures rendered with different tones. Moreover, we benchmark theperformance of two well-known transcription models proposed orig-inally for the piano on this dataset, along with a multi-loss Trans-former model that we newly propose. Our evaluation on this datasetand a separate set of real-world recordings demonstrate the influenceof timbre on the accuracy of guitar sheet transcription, the potentialof using multiple losses for Transformers, as well as the room forfurther improvement for this task.