Samwei1 / CS5647_Text2PianoMusic

CS5647_Text2PianoMusic

Resource code,data and ralated files

Github Link: https://github.com/allisR/CS5647_Text2PianoMusic.git

Train audio decoder

1. Generate dataset. Run preprocess_midi.py in the utils folder. This will write pre-processed data into the data folder split into train, val, and test as per Maestro's recommendation.
2. Run train_decoder.py to train the audio decoder. We provided the pre-trained weight file as all_best_acc.pickle.
3. Test the trained audio decoder by running decoder_generate.py. The output midis should be in the output_midi folder as primer.mid and primer_after.mid.

Train whole model

1. Run train_decoder.py to train the whole model. We provided the model weight file as all_best_acc.pickle.
2. Run model_generate.py to test the whole model and generate a midi file based on the text input. The output midi should be in the output_midi folder as text.mid