musegroup/omr_benchmark

Muse OMR Benchmark

What this is

A small, clean benchmark dataset for OMR (Optical Music Recognition — recognizing music notation from images/PDFs).

It contains 1077 pairs:

• —a symbolic music score (the “ground truth”, see dataset fields below)
• —a corresponding PDF rendering with data augmentation applied

All underlying works are Public Domain.

Why it exists

OMR is often evaluated on private or inconsistent datasets. This dataset aims to provide the community with a practical, reproducible, public benchmark.

What’s inside

Each PDF is generated from our own catalog of PD scores and then augmented to simulate real-world scans:

• —ink blobs / stains
• —scratches / wear
• —crumpled or textured paper
• —rotation / skew
• —other visual noise

Benchmark Code

Check out official repo with evaluation code - https://github.com/musescore/omr_benchmark

Dataset structure

The dataset is distributed as pairs. Typical fields:

• —id: unique sample id
• —pdf_image: augmented PDF file
• —score: symbolic reference in MuseScore Studio file format for evaluation

License

Dataset content is released under CC0-1.0 (no restrictions; attribution appreciated).

Citation

If you use this dataset in a paper or a public benchmark, please cite:

@dataset{pd_omr_benchmark,
  title = {Muse OMR Benchmark},
  author = {Vasily Pereverzev and Kristina Abdullina},
  year = {2025},
}