Optical Music Symbol Recognition (sheet-to-music and music-to-sheet)

Abstract

Optical Music Recognition (OMR) technology aims to transcribe Sheet music into a format that computers can understand and vice versa, allowing Musicians to write, perform, and edit music more easily by simply Capturing a photograph of a music sheet. This study proposes a system that addresses the challenges of existing OMR systems by converting music sheets into a machine-readable format and generating MIDI files. Additionally, the system can convert instrumentation music into musical sheets. The system takes an image of a music sheet as input and uses Convolutional Neural Network to identify symbols, calculate pitch and duration, and output a string corresponding to each line in the music sheet. This string can be played in MIDI or MP3 format. The major steps involved in this process include staff line removal, symbol extraction, symbol recognition, and conversion of musical scores into MIDI format. This study aims to create melodic music that can be performed by humans via a piano interface. The selection of instrument is the piano, which will be the same for both the input and output. Overall, this proposed system has the potential to revolutionize the music industry by making music composition, playing, and editing more accessible to everyone.