-
Developing a Measure Image and Applying It to Deep Learning
- Author(s):
- Lucas Francesco Piccioni Costa, Siqueira Hugo Valadares, Marcella Scoczynski Ribeiro Martins, André Yoshio Caram Ogoshi
- Editor(s):
- Ailynn Ang, Jennifer Bain, David M. Weigl (see profile)
- Date:
- 2023
- Group(s):
- Music Encoding Initiative
- Subject(s):
- Digital humanities, Music
- Item Type:
- Conference poster
- Conf. Title:
- Music Encoding Conference 2022
- Conf. Org.:
- Dalhousie University
- Conf. Loc.:
- Halifax, Nova Scotia, Canada
- Conf. Date:
- May 19-22, 2022
- Tag(s):
- deep learning, Image Codification, Music encoding, Music Theory
- Permanent URL:
- https://doi.org/10.17613/t9dk-2k86
- Abstract:
- The use of intelligent systems linked to musical tasks such as automatic composition, classification, and Music Information Retrieval has increasingly shown itself to be a promising field of study, not only from a computational, but also from a musical point of view. This paper aims to develop an innovative method capable of producing a coded image that contains all the information of a musical measure, generating a structure that can be used in several computational applications involving machine learning, especially deep learning and convolutional neural networks (CNNs). To illustrate the usefulness of this method, the measure image is applied to a CNN to solve the problem of automatic musical harmonization. This brief application achieves better results than those known in the literature, demonstrating the method’s effectiveness.
- Metadata:
- xml
- Status:
- Published
- Last Updated:
- 3 months ago
- License:
- Attribution-NonCommercial-NoDerivatives
- Share this:
Open access, open source, open to all
mec22_122-127_costa.pdf