A machine learning tool to separate earthquake and ambient noise signals for the seismic data in time domain.

1. Install to "WaveDecompNet" virtual envirionment

conda env create -f environment.yml
conda activate WaveDecompNet

2. make a directory called training_datasets in the current folder, download the prepared data from https://www.dropbox.com/s/5frrvx9elzudemt/training_datasets_all_snr_40_unshuffled.hdf5?dl=0 and move it to training_datasets

3. Train the model

python train_model.py

4. A pre-trained model is in folder /Branch_Encoder_Decoder_LSTM and can be directly tested with the downloaded data

python test_model.py

There is a notebook showing a example of how to apply the trained model to continous seismic data of IU.POHA and HV.HAT in the folder ./notebooks/