We provide a deep-learning-based violin fingering generation system that directly exhibits the generated results on the GUI. Given a violin score, our system can generate violin fingering with three models: symbolic, symbolic-audio, and autoregressive. The symbolic model takes the violin score as input and generates a sequence of fingerings at a time. The symbolic-audio model takes both the violin score and the violin audio as inputs, generating fingering with the guidance of the audio. The autoregressive model generates one fingering at a time, taking the violin score and the previously generated fingering as inputs. Users can save the generated fingering as an XML file and edit it with any music notation software.
pip install -r requirements.txt