Researchers have introduced an optical music recognition (OMR) system designed specifically for string quartet scores in PDF format. The work, posted on arXiv, addresses the challenge of digitizing printed sheet music for this ensemble type.
The pipeline starts with a SegNet-based segmentation model that identifies and localizes musical regions within the score pages. After this step, the system applies convolutional processing to extract the musical information, though the abstract does not detail the full architecture.
By focusing on string quartets, the tool targets a specific repertoire where layout and notation conventions may differ from general OMR systems. The approach could streamline the conversion of PDF scores into editable or playable digital formats, though the paper's abstract leaves performance metrics and comparisons to future reading.