WoRMS 2023
5th International Workshop on Reading Music Systems
Milan, Italy, November 4th 2023
All times are given in Central European Time (CET)
9:45 Open Doors
Join the workshop on-site or remotely via Zoom.
10:00 - 11:00 Keynote
Keynote by Werner Goebl (University of Music and Performing Arts Vienna)
Abstract: The final stretch of OMR: Perfecting music score corpora with browser-based editing and validation
The latest optical music recognition (OMR) approaches, benefitting heavily from ever-improving AI models, convert score images to digital score encodings with remarkable accuracy and precision. However, even when error rates might fall to sub-decimal percentages, stakeholders who rely on these digital scores, such as musicians, music scholars or enthusiasts, demand completely error-free material and unwavering trust in the pristinely correct nature of the scores they utilize.
To mitigate this quandary, we propose a decentralized crowd-sourcing system designed to autonomously orchestrate the collective correction of conversion errors in OMR’d digital music score encodings. This system employs a remote-configurable browser-based editing interface named “mei-friend”, capable of juxtaposing specified excerpts of the source score image alongside the corresponding digital rendering. Crucially, this system includes crowd-sourced validation culminating in a distinctive “quality seal” signifying that the correctness of every measure of a given score has received endorsement from multiple individuals.
We hope that this approach represents a significant stride toward realizing a broader vision of cultivating a comprehensive repository of free, open, reliable, easily accessible and FAIR digital scores to be created, edited, validated, and used by a diverse and extensive community of users.
11:00 - 12:30 Session 1
Session chair: Elona Shatri
Optical Music Recognition Workflow for Medieval Music Manuscripts by Ichiro Fujinaga and Gabriel Vigliensoni
The Suzipu Musical Annotation Tool for the Creation of Machine-Readable Datasets of Ancient Chinese Music by Tristan Repolusk and Eduardo Veas
The OmniOMR Project by Jan Hajič, jr., Petr Žabička, Jan Rychtář, Jiří Mayer, Martina Dvořáková, Filip Jebavý, Markéta Vlková, and Pavel Pecina
Towards Music Notation and Lyrics Alignment: Gregorian Chants as Case Study by Juan Carlos Martinez-Sevilla and Francisco J. Castellanos
Session Panel
12:30 - 13:30 Lunch Break
13:30 - 14:45 Session 2
Session chair: Jorge Calvo-Zaragoza
Symbol Generation via Autoencoders for Handwritten Music Synthesis by Jonáš Havelka, Jiří Mayer, and Pavel Pecina
Towards Artificially Generated Handwritten Sheet Music Datasets by Pranjali Hande, Elona Shatri, Benjamin Timms, and George Fazekas
Improving Sheet Music Recognition using Data Augmentation and Image Enhancement by Zihui Zhang, Elona Shatri, and George Fazekas
Session Panel
14:45 - 15:00 Coffee Break
15:00 - 16:15 Session 3
Session chair: Jan Hajič, jr.
Rotations Are All You Need: A Generic Method For End-To-End Optical Music Recognition by Antonio Ríos-Vila
Few-Shot Music Symbol Classification via Self-Supervised Learning and Nearest Neighbor by María Alfaro-Contreras
A Preliminary Study of Few-shot Learning for Layout Analysis of Music Scores by Francisco J. Castellanos, Antonio Javier Gallego, and Ichiro Fujinaga
Session Panel
16:15 - 16:30 Closing Session
Announcements and Goodbye