gpr-gym

Introduction

This repo is an example of how one can generate many training examples with gprMax for use with machine learning models. The config is rough, but it should allow you to get started if you want to mass generate B-scans from gprMax. As outlined in my thesis, generating training data from gprMax on the scale necessary for the training of machine learning models can be difficult and was a major hurdle during my research. Hopefully this will allow you to get started with training data generation for your research. If you end up getting some benefit from this, I would appreciate it if you cite my thesis as one of the Recommended Citations below. Furthermore, feel free to reach out if you have any questions. I will do my best to answer them.

Installation Instructions

git clone --recursive [email protected]:will-rice/gpr-gym.git
cd gprMax
python setup.py build
python setup.py install
cd ..
pip install -r requirements.txt

Generate Training Data

python generate_training_data.py

Recommended Citations

This citation is recommended by the university

Rice, William, "Applying generative adversarial networks to intelligent subsurface imaging and identification" (2019). Masters Theses and Doctoral Dissertations.
<https://scholar.utc.edu/theses/595>

However, here is the arxiv bibtex if you prefer.

@ARTICLE{2019arXiv190513321R,
       author = {{Rice}, William},
        title = "{Applying Generative Adversarial Networks to Intelligent Subsurface Imaging and Identification}",
      journal = {arXiv e-prints},
     keywords = {Electrical Engineering and Systems Science - Image and Video Processing, Computer Science - Machine Learning},
         year = "2019",
        month = "May",
          eid = {arXiv:1905.13321},
        pages = {arXiv:1905.13321},
archivePrefix = {arXiv},
       eprint = {1905.13321},
 primaryClass = {eess.IV},
       adsurl = {<https://ui.adsabs.harvard.edu/abs/2019arXiv190513321R}>,
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}