This repository contains the code of the paper accepted and to be published in the proceeding of the EvoMusArt 2021, the 10th International Conference on Artificial Intelligence in Music, Sound, Art and Design http://www.evostar.org/2021/evomusart/
If you are using this code in full or in part, please cite our work as
@inproceedings{EvoMusArt2021,
author = {F. Marchetti and C. Wilson and C. Powell and E. Minisci and A. Riccardi},
editor = {},
title = {Convolutional Generative Adversarial Network,via Transfer Learning, for Traditional ScottishMusic Generation},
booktitle = {Computational Intelligence in Music, Sound, Art and Design - 10th International
Conference, EvoMUSART 2021, Held as Part of EvoStar 2021, Seville,
Spain, April, 2021, Proceedings},
series = {Lecture Notes in Computer Science},
volume = {},
pages = {},
publisher = {Springer},
year = {2021}
}
The paper presents an application of the Binary Multi-track Sequential Generative Adversarial Network (BinaryMuseGAN) to generate original Scottish music from a reduced dataset of songs by exploiting a Transfer Learning approach. The GAN model is first trained on a larger and diverse music collection, to learn representative features of music in general, and then fine tuned on the smaller dataset of traditional Scottish music.
The work has been completed by the researchers of the Intelligent Computational Engineering Laboratory (ICE Lab) of the Univeristy of Strathclyde during Covid 2020 lockdown.