Beat this! Accurate beat tracking without DBN postprocessing

TL;DR

This paper introduces a beat tracking system that achieves high accuracy and generality across diverse music genres without using DBN postprocessing, by employing a new loss function and a hybrid convolution-transformer architecture.

Contribution

The authors develop a novel beat tracking system trained on multiple datasets that removes the need for DBN postprocessing, improving accuracy and generality.

Findings

Surpasses current state-of-the-art F1 score without DBN.

Effective across diverse music genres and tempo variations.

Published code and datasets for community use.

Abstract

We propose a system for tracking beats and downbeats with two objectives: generality across a diverse music range, and high accuracy. We achieve generality by training on multiple datasets -- including solo instrument recordings, pieces with time signature changes, and classical music with high tempo variations -- and by removing the commonly used Dynamic Bayesian Network (DBN) postprocessing, which introduces constraints on the meter and tempo. For high accuracy, among other improvements, we develop a loss function tolerant to small time shifts of annotations, and an architecture alternating convolutions with transformers either over frequency or time. Our system surpasses the current state of the art in F1 score despite using no DBN. However, it can still fail, especially for difficult and underrepresented genres, and performs worse on continuity metrics, so we publish our model, code, and preprocessed datasets, and invite others to beat this.