For over 50 years, researchers have been trying to teach computers to read music notation, referred to as Optical Music Recognition (OMR). However, this field is still difficult to access for new researchers, especially those without a significant musical background: Few introductory materials are available, and, furthermore, the field has struggled with defining itself and building a shared terminology. In this work, we address these shortcomings by (1) providing a robust definition of OMR and its relationship to related fields, (2) analyzing how OMR inverts the music encoding process to recover the musical notation and the musical semantics from documents, and (3) proposing a taxonomy of OMR, with most notably a novel taxonomy of applications. Additionally, we discuss how deep learning affects modern OMR research, as opposed to the traditional pipeline. Based on this work, the reader should be able to attain a basic understanding of OMR: its objectives, its inherent structure, its relationship to other fields, the state of the art, and the research opportunities it affords.

Optical Music Recognition (OMR) is the challenge of understanding the content of musical scores. Accurate detection of individual music objects is a critical step in processing musical documents, because a failure at this stage corrupts any further processing. So far, all proposed methods were either limited to typeset music scores or were built to detect only a subset of the available classes of music symbols. In this work, we propose an end-to-end trainable object detector for music symbols that is capable of detecting almost the full vocabulary of modern music notation in handwritten music scores. By training deep convolutional neural networks on the recently released MUSCIMA++ dataset which has symbol-level annotations, we show that a machine learning approach can be used to accurately detect music objects with a mean average precision of up to 80%.

Deep learning is bringing breakthroughs to many computer vision subfields including Optical Music Recognition (OMR), which has seen a series of improvements to musical symbol detection achieved by using generic deep learning models. However, so far, each such proposal has been based on a specific dataset and different evaluation criteria, which made it difficult to quantify the new deep learning-based state-of-the-art and assess the relative merits of these detection models on music scores. In this paper, a baseline for general detection of musical symbols with deep learning is presented. We consider three datasets of heterogeneous typology but with the same annotation format, three neural models of different nature, and establish their performance in terms of a common evaluation standard. The experimental results confirm that the direct music object detection with deep learning is indeed promising, but at the same time illustrates some of the domain-specific shortcomings of the general detectors. A qualitative comparison then suggests avenues for OMR improvement, based both on properties of the detection model and how the datasets are defined. To the best of our knowledge, this is the first time that competing music object detection systems from the machine learning paradigm are directly compared to each other. We hope that this work will serve as a reference to measure the progress of future developments of OMR in music object detection.

No abstract

No abstract