The objective of the project is Neural Style Transfer done on audio files. The project included the implementation of the StarGAN neural network architecture, in order to acchive many-to-many style transfer, applied on sound spectrograms. The neural network consisted of mostly convolutional generator, discriminator, as well as the pretrained mainly convolutional classifier.
Reference paper: https://arxiv.org/pdf/1806.02169.pdf git: https://github.com/liusongxiang/StarGAN-Voice-Conversion.git
Dataset: https://www.kaggle.com/andradaolteanu/gtzan-dataset-music-genre-classification