This is the work with this paper in IEEE International Conference on Semantic Computing ICSC 2020: link
We implement EC2-VAE into the conditional generative model to let people generate the music melody in terms of controlling rhythm patterns and chord progressions, and even extra chord function labels.
- processed_data: processed Nottingham data in EC2-VAE latent vector sequences, due to the 100MB limits, we have some missing files here.
- vae: EC2-VAE model
- AmMGM_model_decode.ipynb: about how to use the trained model parameters to generate the music from train/valid/test dataset.
- model_mask_cond: conditional generative model
- train_AmMGM: training model file.
- result: the vae_nottingham_output, model_generation_out, and sample_for_presentation
We did not provide the trained parameters in github, if you want find out both AmMGM-parameters and EC2-VAE-parameters we trained for this model, check out the link here.
Please cite this paper if you want to base on this work to make improvements or further research.
@inproceedings{amg-ec2vae-icsc,
author = {Ke Chen and Gus Xia and Shlomo Dubnov},
title = {Continuous Melody Generation via Disentangled Short-Term Representations and Structural Conditions},
booktitle = {{IEEE} 14th International Conference on Semantic Computing, {ICSC}},
pages = {128--135},
publisher = {{IEEE}},
year = {2020},
address = {San Diego, CA, USA}
}