OpenScore String Quartets: Scanned and Rendered Score Images with MusicXML Ground Truth

Guangyangmusic derived this dataset from the OpenScore String Quartets corpus for evaluating Optical Music Recognition systems. It contains a subset of string quartets from the 'long 19th century' with both scanned images of real scores and corresponding MusicXML ground truth, plus clean images rendered from the MusicXML. The dataset was last updated on Hugging Face in April 2026.

Use Cases

• Train OMR models based on the paired scanned images and MusicXML ground truth.
• Benchmark OMR system accuracy based on the provided evaluation corpus.
• Compare OMR performance on real scanned scores versus clean rendered images.
• Analyze music notation patterns in 19th-century string quartets based on the MusicXML data.

Strengths

• Contains paired data types: scanned score images, MusicXML ground truth, and rendered clean images.
• Specifically designed for the concrete task of Optical Music Recognition system evaluation.
• Derived from a published corpus (Gotham et al., 2023) focused on a specific musical period and ensemble.

Limitations

• Row count, column names, and file formats are unknown, limiting suitability assessment.
• The dataset is a subset of a larger corpus; the selection criteria and representativeness are not detailed.
• License information is absent, which may restrict usage.

Provenance

Source

OpenScore String Quartets corpus (Gotham et al., 2023)

Collection Method

Subset extracted containing scanned images and MusicXML, with additional images rendered via MuseScore.

Time Range

Music from the 'long 19th century'.

Freshness

Last updated 2026-04-03 22:54:30.

Geography

null