ChopinBotApp

Web application for piano music generation in the style of Chopin. Powered by Streamlit.io

Currently deployed at https://share.streamlit.io/msgilmer/chopinbotapp/main/ChopinBotApp.py

The ML model used by this application was developed in https://github.com/msgilmer/Springboard_Capstone. Specifically, in notebooks/model_training-v3.ipynb.

Following Flowchart created with LucidChart:

</br></br> <b>Directory Structure</b>

.
├── ChopinBotApp.py               # The script that is deployed.
├── ChopinBotApp_local.py         # The local version (run as streamlit run ChopinBotApp_local.py).
├── X_val                       
│   ├── X_val_ext_1.npy           # The X part of the validation set broken into pieces to avoid git-lfs (not currently compatible with Streamlit). Copied from msgilmer/Springboard_Capstone/train_and_val.
│   ├── booleanized_sequences.npy # The same validation set broken up into the binary classification part (saved with dtype = np.bool) and the duration part. Generated from the above two via ../separate_X_val.py.
│   └── durations.npy
├── images                        # Contents copied from msgilmer/Springboard_Capstone/images. Images posted to application.
│   ├── chopin.jpg    
│   ├── precision_and_recall.jpg
├── logfiles                      # Logfiles for each unique session. Used for monitoring errors.
│   └── logzero*log
├── midi_output                   # Directory to store generated MIDI files for download for the local version only.
│   ├── generated.mid
│   └── seed.mid
├── models                        # Directory to store pre-trained keras model weights, copied from msgilmer/Springboard_Capstone/models.
│   └── best_maestro_model_weights_batchnorm_p_0_ext20_2_1_512_0pt3_mnv_2.h5
├── requirements.txt              # Required to deploy the app. Generated with pipreqs.
├── rndm_seed_index_files         # Text files that uniquely store the last randomly-generated seed index (for the X_val set) for each unique session.
│   └── rndm*txt
├── separate_X_val.py             # Script used to save the X_val set into a more energy-efficient format (X_val/booleanized_sequences.npy and X_val/durations.npy) for consuption by ChopinBotApp*py.
└── ...

│   ├── X_val_ext_2.npy