Abstract
Optical music recognition (OMR) is an area in music information retrieval. Music object detection is a key part of the OMR pipeline. Notes are used to record pitch and duration and have semantic information. Therefore, note recognition is the core and key aspect of music score recognition. This paper proposes an end-to-end detection model based on a deep convolutional neural network and feature fusion. This model is able to directly process the entire image and then output the symbol categories and the pitch and duration of notes. We show a state-of-the-art recognition model for general music symbols which can get 0.92 duration accurary and 0.96 pitch accuracy .
Highlights
Today, many pieces of music are recorded and passed down by scores
The results show that our model performs very well on score recognition tasks
We proposed a complete end-to-end printed score recognition model based on a deep convolutional neural network
Summary
Many pieces of music are recorded and passed down by scores. Musical scores play a decisive role in the development of music. Music has been preserved in the form of pictures, whether the composer’s manuscript or a published version. The storage of music scores in picture form has brought plenty of difficulties for Music Information Retrieval (MIR) [1]. Music scores ought to be recorded in digital format, which is unique to music, rather than pixels, which are unrelated to each other, to make it easy for people to retrieve or edit it. To edit a score stored in a picture, we have to manually enter elements one by one into music notation software before making changes and adjustments. There are many existing music-encoding formats including
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
