BGT-G2G

Abstract:

With the rapid development of hardware, software and algorithms, the use of artificial intelligence for artistic creation has become a reality, especially in music and image creation. Examples include AI composition using recurrent neural networks, generative adversarial networks, etc., and image style migration using convolutional neural networks and their relatives. In existing works, good results can already be achieved for single-form art compositions. However, little is known about the integration of art across forms. This article starts with AI composition and proposes a state-of-the-art model that uses images as content input and music as style input to achieve cross-form style migration from images to music. Further contributions has been made in image pre-processing, and the training of a music style migration model. The findings of this project provide further ideas and inspiration for the creation of multi-media art.

For experimental results, please see here.