vix.ing · top · new · best · stats · spec

Detecting Noteheads in Handwritten Scores with ConvNets and Bounding Box\n Regression

2017/08/05 by Jan Hajič, Hajič, Jan, Pavel Pecina +1 · 2 citations
Arts and Humanities · Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #Diverse Musicological Studies #FOS: Computer and information sciences #Handwritten Text Recognition Techniques #I.4.6 #I.5.1 #I.5.4 #I.7.5 #Music Technology and Sound Studies #Music and Audio Processing

paper · pdf · doi:10.48550/arxiv.1708.01806

openalex publication_date 2017/08/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Noteheads are the interface between the written score and music. Each\nnotehead on the page signifies one note to be played, and detecting noteheads\nis thus an unavoidable step for Optical Music Recognition. Noteheads are\nclearly distinct objects, however, the variety of music notation handwriting\nmakes noteheads harder to identify, and while handwritten music notation symbol\n em classification is a well-studied task, symbol em detection has usually\nbeen limited to heuristics and rule-based systems instead of machine learning\nmethods better suited to deal with the uncertainties in handwriting. We present\nongoing work on a simple notehead detector using convolutional neural networks\nfor pixel classification and bounding box regression that achieves a detection\nf-score of 0.97 on binary score images in the MUSCIMA++ dataset, does not\nrequire staff removal, and is applicable to a variety of handwriting styles and\nlevels of musical complexity.\n

Citations

Cited by

Related