Máster Universitario en Ciencia de Datos
In this Master’s thesis we will produce a complete music recognition system based on two steps: an image-agnostic transcription of an input staff image and an automatic translation step between the output of such recognition (which is in such an agnostic encoding) to a semantic encoding. First, we will investigate more about how to approach the problem of machine translation between musical languages, as this is something that has not yet been explored in the existing literature. We will propose various systems, from statistical to deep neural network based, to tackle the problem and evaluate their performance on different datasets. Once this step is done, we will evaluate the feasibility of these systems and propose the different situations in which an automatic translator between two music encodings may be suitable.
Technical report (in spanish) at UA repostirory: