Deep-learning based ocean bottom seismic wavefield recovery (PyTorch)

Codes for generating results in Siahkoohi, A., Kumar, R. and Herrmann, F.J., 2019. Deep-learning based ocean bottom seismic wavefield recovery. In SEG Technical Program Expanded Abstracts 2019 (pp. 2232-2237). doi: 10.1190/segam2019-3216632.1.

Prerequisites

This code has been tested using Deep Learning AMI (Amazon Linux) Version 24.2 on Amazon Web Services (AWS). We performed the test on g3s.xlarge and g4dn.xlarge instances. Follow the steps below to install the necessary libraries:

cd $HOME
git clone git@github.com/alisiahkoohi/wavefield-reconstruction-torch.git
cd $HOME/wavefield-reconstruction-torch
conda create -n torch pip python=3.6
source activate torch
conda install pytorch torchvision cudatoolkit=10.1 -c pytorch #If your system has GPU
pip install --user -r  requirements.txt

Dataset

Links have been provided in RunTraining.sh script to automatically download the 14.33 Hz monochromatic seismic data into the necessary directory. Total size of the dataset for each fequency is 6.52GB + 6.52GB + 6.52GB + 118KB.

Script descriptions

RunTraining.sh: script for running training. It will make model/ and data/ directory in $HOME for storing training/testing data and saved neural net checkpoints and final results, respectively. Next, it will train a neural net for the experiment for 14.33 Hz monochromatic seismic data.

RunTesting.sh: script for testing the trained neural net. It will reconstruct the entire subsampled 14.33 Hz monochromatic seismic data and place the result in sample/ directory to be used for plotting purposes.

src/main.py: constructs wavefield_reconstrcution class using given arguments in RunTraining.sh, defined in model.py and calls train function in the defined wavefield_reconstrcution class.

src/model.py: includes wavefield_reconstrcution class definition, which involves train and test functions.

Running the code

To perform training, run:

# Running on GPU

bash RunTraining.sh

To evaluated the trained network on test data set run the following. It will automatically load the latest checkpoint saved.

# Running on GPU

bash RunTesting.sh

To generate and save figures shown in paper for 14.33 Hz monochromatic seismic data run the following:

bash utilities/genFigures.sh

The saving directory can be changed by modifying savePath variable in utilities/genFigures.sh.

Citation

If you find this software useful in your research, please cite:

@conference {siahkoohi2019SEGdlb,
	title = {Deep-learning based ocean bottom seismic wavefield recovery},
	booktitle = {SEG Technical Program Expanded Abstracts},
	year = {2019},
	note = {(SEG, San Antonio)},
	month = {09},
	pages = {2232-2237},
	keywords = {machine learning, obn, reciprocity, reconstruction, SEG},
	doi = {10.1190/segam2019-3216632.1},
	author = {Ali Siahkoohi and Rajiv Kumar and Felix J. Herrmann}
}

Questions

Please contact alisk@gatech.edu for further questions.

Author

Ali Siahkoohi