DeepISMNet: using synthetic datasets to train an end-to-end Convolutional Neural Network for implicit structural modeling

Three-Dimensional Implicit Structural Modeling Using Convolutional Neural Network

**This is a Pytorch version of a deep learning method using a convolution neural network to predict a scalar field from sparse structural data associated with multiple distinct stratigraphic layers and faults.

As described in Three-Dimensional Implicit Structural Modeling Using Convolutional Neural Network by Zhengfa Bi¹, Xinming Wu¹, Zhaoliang Li², Dekuan Change³ and Xueshan Yong³. ¹University of Science and Technology of China; ²China Aero Geophysical Survey and Remote Sensing Center for Natural Resources; ³Research Institute of Petroleum Exploration & Development-NorthWest(NWGI), PetroChina.

Requirments

python>=3.6
torch>=1.0.0
torchvision
torchsummary
natsort
numpy
pillow
plotly
pyparsing
scipy
scikit-image
sklearn
tqdm

Install all dependent libraries:

pip install -r requirements.txt

Dataset

To train our CNN, we automatically created numerous structrual models and the associated data with distinct stratigraphic layers and faults, which were shown to be sufficient to obtain an excellent structural modeling network.

The synthetic structural models can be downloaded from here, and the input data are randomly generated in training.

Run training_data_generating.ipynb to create a new synthetic dataset.

Training

Run train.ipynb to start training a new DeepISMNet model by using the synthetic dataset.

License

This extension to the Pytorch library is released under a creative commons license which allows for personal and research use only. For a commercial license please contact the authors. You can view a license summary here: http://creativecommons.org/licenses/by-nc/4.0/