Summary
Optical Music Recognition (OMR) and Automatic Music Transcription (AMT) are the research fields that investigate how to computationally transcribe music score images and audio recordings, respectively, into digital scores. After decades of research, neither AMT nor OMR have lived up to their commitment and remain open challenges, with plenty of room for improvement. MultiScore seeks to unlock the current situation by levering vast amounts of annotated data to apply state-of-the-art technologies in deep neural networks, and also to find intersections and synergies in both research lines that have previously been addressed separately.