Antonio Ríos-Vila

A Transformer Approach for Polyphonic Audio-to-Score Transcription

End-to-end Audio-to-Score (A2S) transcription aims to derive a score that represents the music content of an audio recording in a single step. While current state-of-the-art methods, which rely on Convolutional Recurrent Neural Networks trained with the Connectionist Temporal Classification loss fun…