Problem
Research questions and friction points this paper is trying to address.
Optical Music Recognition
String Quartet
Multi-part Score
Dataset
Music Transcription
A Dataset and Benchmark for Optical Music Recognition of String Quartet Scores
🤖 AI Summary
This work addresses the limited progress in multi-stave optical music recognition (OMR) due to the absence of dedicated datasets and standardized benchmarks. To bridge this gap, we introduce OSSQ-OMR, the first multi-stave OMR dataset tailored for string quartets, derived from the OpenScore corpus. Leveraging image–score alignment techniques, OSSQ-OMR provides both system-level and staff-level images accompanied by precise transcriptions in three symbolic formats: LMXE, **kern, and ABC. We also establish a standardized evaluation protocol and implement LSTM and Transformer baseline models. Experimental results show that these baselines achieve an OMR-NED of 3.6% on synthetic images and 5.9% on scanned images. Notably, the LSTM exhibits significantly less performance degradation on real scanned images—approximately 1/2.6 that of the Transformer—highlighting its superior robustness in practical scenarios.
📝 Abstract
Optical music recognition (OMR) transcribes music scores into digital formats. While the field has advanced significantly on monophonic and piano-form scores, multi-part score transcription remains underexplored, largely due to the absence of a suitable dataset. We introduce OpenScore String Quartet for Optical Music Recognition (OSSQ-OMR), the first dataset dedicated to multi-part OMR. Built on the OpenScore String Quartet corpus, OSSQ-OMR pairs digitally encoded scores with their original scanned editions from IMSLP, with all images visually aligned to their transcriptions. The dataset is released with score images at system and staff levels, and paired transcriptions in three encoding formats: Extended Linearized MusicXML (LMXE), **kern, and ABC. In total, OSSQ-OMR contains 24,544 system images and 98,172 staff images drawn from 116 string quartet scores. We accompany the dataset with a benchmark protocol and baseline results from two representative OMR models, evaluated across four random score-level splits with mutually exclusive test sets. Baselines reach OMR-NED as low as 3.6% on synthetic and 5.9% on scanned inputs; results reveal substantial effects of encoding and segmentation choices, with the LSTM-based baseline degrading on scanned inputs roughly 2.6 times less than the Transformer-based baseline.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Optical Music Recognition
String Quartet
Multi-part Score
Dataset Benchmark
Music Transcription
🔎 Similar Papers
No similar papers found.
