On 21 June, Music Day, let's take a look at computers and music...
Intuidoc's research focuses on handwriting, gesture (2D and 3D) and document processing from a number of angles: analysis, recognition, composition and interpretation. Within the team, Bertrand Coüasnon, associate professor at INSA Rennes, and Aurélie Lemaitre, associate at Université Rennes 2, are particularly interested in the recognition of images of old musical scores.
Recognising and transcribing
Based on deap learning and artificial intelligence techniques, musical score recognition consists of analysing, recognising and interpreting the scanned document in order to transcribe it as reliably as possible, so that the score can then be re-edited collaboratively, synchronised with a concert recording, translated into Braille or interpreted.
Trained to suggest possible solutions
Each musical symbol (notes, note values, bars, etc.) will be analysed to extract its meaning, based on musical knowledge. Trained by deep learning techniques, the system provides reliable recognition of the score and can even detect potential errors linked to partially erased musical symbols and suggest corrective hypotheses.
This work is being carried out as part of the ANR CollabScore project, led by the CNAM, with partners IReMus, IRISA, the University of Lille, the BnF and the Royaumont Foundation.
To find out more about the work of the Intuidoc team:
credit photo : pexels ylanite Koppens