MusiCorpus v1.0

The MusiCorpus v1.0 dataset provides 1,309 pages of historical sheet music, primarily handwritten, with MusicXML transcriptions and symbol annotations, for training and evaluating Optical Music Recognition (OMR) systems in realistic conditions. It is the largest dataset of handwritten music to date and the first dataset containing a realistic representative sample of musical document collections from memory institutions. A large amount of musical heritage has been digitised by archival institutions, but the field of Optical Music Recognition (OMR) has struggled with making this music machine-readable and therefore findable despite progress in deep learning methods, because no datasets for training systems in realistic conditions were available. MusiCorpus is suitable for training and evaluating both end-to-end systems and object detection-based OMR systems and comparing their performance.