Neural Network for generation of lo-fi music for CS486. There is two separated sections for CNN and RNN generation methods.
Follow this steps to run the project:
- In the .ipynb file run code by blocks.
- Run "Processing the midi into a notes array" blocks. This only needs to be run once to store the notes as an array to the file.
- Run the "Processing the notes array into the numpy array format for the NN" block.
- Run "Create the model and its layers" which will build the specified model (default one is the one tuned by us).
- If you want to train the model and store weights from scratch load previously stored weights into the model, run "Generating model from previously trained weights" with your desired file path.
- After training or loading weights into a model, run "Generate Music with Model" to produce some notes/chords and then run "Play the midi" to listen to the results.