project renewal is in progress.
Keras implementation of "Chord Generation from Symbolic Melody Using BLSTM Networks"
- Piano Melody Sound : MuseScore2
- Guitar Backing Sound : I play
- Python 3.6
- Numpy
- Pandas
- Keras 2.2.4
data_preprocess.py
- Transpose to the C key
- Convert to 2 kinds of chords (maj or min)
make_npy.py
- Create numpy array file for training
train.py
- Feed input/target vector into Neural net
- Train the model
generate.py
- Generate the Chord from note sequence
-
Neural Network Algorithm
I compared Attention LSTM with other RNN algorithms(LSTM, BLSTM, GRU), but there was no performance difference. -
Some hyper parameters
-
Preprocess method
I didn't use the feature about note's time information(time, duration).
I used the dataset provided in the paper, the dataset is based on the Wikifonia Dataset.