This repository makes use of the CSV dataset provided in: https://www.kaggle.com/saikayala/jazz-ml-ready-midi and acts as completion for Assignment #05 of Bangkit Academy.
We use the first 100 MIDI songs to train a LSTM neural network to generate a MIDI song containing 500 notes.
This project makes use of the following libraries:
- NumPy
- pandas
- Pickle
- Tensorflow
- Keras
- scikit-learn
- Music21
- Jupyter Notebook
Run model.ipynb to train the model from the dataset jazz-midi.csv. The notes files are saved at x.pickle and notes.pickle, while the model and weights are saved at model.json and model.h5 respectively.
To generate a MIDI file from the saved notes and model, run predict.ipynb, which will save the output MIDI at output.mid.