This is the official implementation of MusER (AAAI'24), which employs musical element-based regularization in the latent space to disentangle distinct musical elements, investigate their roles in distinguishing emotions, and further manipulate elements to alter musical emotions.
- Paper link
- Check our demo page and listen!
The dependency packages can be found in requirements.txt file. One can use pip install -r requirements.txt to configure the environment. We use python 3.8 under Ubuntu 20.04.6 LTS to run the experiments. We highly recommend using the conda environment for deployment.
1.First, train the MusER, use the following command:
python MusER.py --data_path ./data/co-representation/emopia_data.npz --dataset emopia --model_path your_model_saving_path --log_path your_log_path
Explanations on the parameters:
data_path: training data path.
dataset: which dataset to train, choices include emopia and ailabs and should be consistent with data_path.
2.Then, run the VQ_explore.py to get VQ_dict.data (store codebook discrete indexes that are actually used during training) for training prior model and to visulize the latent space:
python VQ_explore.py --data_path ./data/co-representation/emopia_data.npz --VQ_VAE MusER_TRANS_CA_GE_emopia --plot_flag True
Explanations on the parameters:
VQ-VAE: the name of the trained VQ-VAE model, such as MusER_TRANS_CA_GE_emopia.
plot_flag: whether to plot latent space visualization map, boolean variable.
3.Next, train the conditional autoregressive prior model:
python Prior_train.py --data_path ./data/co-representation/emopia_data.npz --dataset emopia --VQ_VAE MusER_TRANS_CA_GE_emopia --model_path your_prior_model_saving_path --log_path your_log_path
4.Finally, generate music using your trained MusER and Prior model:
python generate.py --VQ_prior Prior_TRANS_CA_ER_emopia --VQ_VAE MusER_TRANS_CA_GE_emopia --music_file your_music_file_name
Explanations on the parameters:
VQ_prior: the name of the trained prior model, such as Prior_TRANS_CA_GE_emopia.
VQ-VAE: the name of the trained VQ-VAE model, such as MusER_TRANS_CA_GE_emopia.
music_file: the name of the file that stores music generated by VQ-VAE and VQ_prior.
If you find the code useful for your research, please consider citing
@inproceedings{ji2024muser,
title={MusER: Musical Element-Based Regularization for Generating Symbolic Music with Emotion},
author={Ji, Shulei and Yang, Xinyu},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={38},
number={11},
pages={12821--12829},
year={2024}
}