An Empirical Evaluation of End-to-End Polyphonic Optical Music Recognition

TL;DR

This paper introduces new datasets and models for end-to-end polyphonic optical music recognition, achieving state-of-the-art results by treating the task as multi-sequence detection with novel decoder architectures.

Contribution

It presents two innovative formulations for polyphonic OMR and introduces the RNNDecoder, improving recognition accuracy on complex polyphonic scores.

Findings

RNNDecoder achieves state-of-the-art performance.

New datasets enable large-scale polyphonic OMR evaluation.

Multi-sequence detection outperforms previous methods.

Abstract

Previous work has shown that neural architectures are able to perform optical music recognition (OMR) on monophonic and homophonic music with high accuracy. However, piano and orchestral scores frequently exhibit polyphonic passages, which add a second dimension to the task. Monophonic and homophonic music can be described as homorhythmic, or having a single musical rhythm. Polyphonic music, on the other hand, can be seen as having multiple rhythmic sequences, or voices, concurrently. We first introduce a workflow for creating large-scale polyphonic datasets suitable for end-to-end recognition from sheet music publicly available on the MuseScore forum. We then propose two novel formulations for end-to-end polyphonic OMR -- one treating the problem as a type of multi-task binary classification, and the other treating it as multi-sequence detection. Building upon the encoder-decoder architecture and an image encoder proposed in past work on end-to-end OMR, we propose two novel decoder models -- FlagDecoder and RNNDecoder -- that correspond to the two formulations. Finally, we compare the empirical performance of these end-to-end approaches to polyphonic OMR and observe a new state-of-the-art performance with our multi-sequence detection decoder, RNNDecoder.