Towards an Integration of OMR Models in Pattern Improvisation

In this paper, we present a prototype for an interactive musical system, which allows the user to manage musical patterns in real-time using images of printed music in traditional Western notation. Monophonic scores are interpreted by a pre-trained Optical Music Recognition model, converted into numeric sequences, and sent to a pattern generator/synthesizer. This prototype constitutes a novel proposal, integrating deep learning and computer vision techniques into music making and promoting interactivity across disciplines.