Automatic Music Transcription
CNN training codes retrieved from here. DNN/LSTM/BiLSTM codes retrieved from here.
Install dependencies: pip install numpy tensorflow streamlit librosa h5py pandas matplotlib plotly numba scipy SoundFile
Put model files in /sl_data/models/{modelName}_{nEpoch}_{nEarly} folder. This includes models and loss logs (model_{modelName}_{fold}.h5 and losses_log_{fold}.h5).
Put music files in /sl_data/presets folder. This comes in the form of .wav files and optionally .txt and .mid (for CNN) ground truth files with the same name.
Put standardization or normalization boundaries in /sl_data/std folder. The files should be named as means_stds.h5, mean_stds_nm, and minmax_meanlabel.h5
*losses_log and means files are custom, tweaking might be needed to suit you.
Run sl.py via streamlit:
streamlit run sl.py