Neural Networks 2020

Project repository for the course of Neural Networks, Sapienza University of Rome.

Analysis and Comparison of Deep Learning Methods for Jazz Music Generation.

This repository contains notebooks about the implementation of the WaveNet, Simplified_WaveNet and LSTMNet. In addition, it contains libraries of methods for data pre and post processsing, datasets analysis, a notebook to use Magenta models and a novel dataset of jazz MIDI files.

Contents:

• Structure
• Execution
• References
• Team Members

Structure

• audio_samples folder contains audio tracks generated by the five models and extracted from the dataset
• novel_dataset is a novel dataset of well-known jazz songs in MIDI extension
• scripts folder contains all the scripts for pre&post processing, datasets analysis and the notebooks for the networks
• presentation.pdf is the presentation of this project
• report.pdf is the report for this project

Execution

WaveNet

Run the scripts/WaveNet_Notebook.ipynb notebook.

Simplified_WaveNet

Run the scripts/SimplifiedWaveNet_Notebook.ipynb notebook.

LSTMNet

Run the scripts/LSTMNet_Notebook.ipynb notebook.

GRUNet

1. Clone the repository in https://github.com/haryoa/note_music_generator.git.
2. Move the MusicGeneratorPianoColab.ipynb in the same directory of your dataset.
3. Customize the notebook changing filepaths with correct values.
4. Run the notebook

Attention_RNN

1. Clone the repository in https://github.com/magenta/magenta.git.
2. Move the Attention_RNN_Notebook.ipynb in the same directory.
3. Run the notebook.

References

• A. Van Den Oord, S. Dieleman, H. Zen, K. Simonyan, O. Vinyals, A. Graves, N. Kalchbrenner, A. Senior, K. Kavukcuoglu. WaveNet: A Generative Model for Raw Audio. Proc. 9th ISCA Speech Synthesis Workshop, 2016, pp. 125-125.
• A. Géron. Hands-On Machine Learning with Scikit-Learn, Keras and TensorFlow. 2019, O’Reilly, 2nd Edition, pp. 521-523.
• J. M. Simões, P. L. Machado, A. C. Rodrigues. Deep Learning for Expressive Music Generation. In Proceedings of the 9th International Conference on Digital and Interactive Arts (ARTECH 2019). Association for Computing Machinery, New York, NY, USA, Article 14, 1–9. DOI:https://doi.org/10.1145/3359852.3359898.
• https://github.com/haryoa/note_music_generator
• https://github.com/magenta/magenta/tree/master/magenta/models/melody_rnn
• https://www.kaggle.com/saikayala/jazz-ml-ready-midi
• A. Huang, R. Wu. Deep Learning for Music. 2016, arxiv: http://arxiv.org/abs/1606.04930.
• H. W. Dong, W. Y. Hsiao, L. C. Yang, Y. H. Yang. MuseGAN: Multi-Track Sequential Generative Adversarial Networks for Symbolic Music Generation and Accompaniment. In Thirty-Second AAAI Conference on Artificial Intelligence, 2018.
• H. Hung, C. Wang, Y. Yang and H. Wang, Improving Automatic Jazz Melody Generation by Transfer Learning Techniques., 2019 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), Lanzhou, China, 2019, pp. 339-346, doi: 10.1109/APSIPAASC47483.2019.9023224.

Team members

• Ratini Riccardo, 1656801.
• Simionato Giada, 1822614.