Researchers have introduced OSSQ-OMR, the first dataset specifically designed for optical music recognition (OMR) of multi-part scores, particularly string quartets. This dataset, derived from the OpenScore String Quartet corpus and IMSLP, pairs digitally encoded music transcriptions with their scanned editions. OSSQ-OMR includes over 24,000 system images and nearly 100,000 staff images from 116 string quartet scores, with transcriptions available in LMXE, kern, and ABC formats. The release also features a benchmark protocol and baseline results from LSTM and Transformer-based models, highlighting encoding and segmentation impacts on performance. AI
IMPACT This dataset and benchmark could accelerate research and development in optical music recognition, particularly for complex, multi-part musical scores.
RANK_REASON The item describes a new dataset and benchmark protocol for optical music recognition, which falls under academic research. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
- Extended Linearized MusicXML
- International Music Score Library Project
- kern
- LMXE
- long short-term memory
- OMR-NED
- OpenScore String Quartet
- OpenScore String Quartet for Optical Music Recognition
- OSSQ-OMR
- Transformer++
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
