GeoFlow

Dataset management interface with Keras for geophysics

Contributing

This repository is organized in the following fashion. From highest level to lowest:

• The file Dataset2Dtest.py shows an example of how to generate a training set, and train a NN with it. This is the starting point.
• A neural network is defined in RCNN2D.py. This class builds the NN and the loss. It is used in Dataset2Dtest.py. To build a new network, a child class can be defined from RCNN2D.
• To help with training, a class Trainer is provided. It needs a GeoDataset class and a RCNN2D-like class as input.
• The file Dataset_define.py is where different Datasets are defined.
• A GeoDataset is implemented with the GeoDataset class. It provides an interface to generate 2D velocity models and model the seismic data with fixed parameters.
• The GeoDataset class contains the method set_dataset. A new GeoDataset can be defined by defining a child class from the GeoDataset base class and overriding set_dataset. This method needs to return three objects based on three classes
  • EarthModel. This class allows to generate a random model. It is based the ModGen library available upon request. Different model generator can be defined from this case (see MarineModel).
  • Acquisition defines all the parameters for the creation of the seismic data by SeisCL. In particular, override the method set_rec_src to define a different acquisition setup.
  • OutputGenerator is a class that generate the labels from the model and acquires objects.

Style guide

Code style should follow closely PEP 8. Commits should follow closely git good practices. Refer to STYLEGUIDE.md for a comprehensive guide to this project's style conventions.

Installation

IMPORTANT: Use the latest SeisCL version on the devel branch.

You should clone this repository

git clone https://github.com/gfabieno/SeisCL.git

a) Use Docker (easiest)

We provide a Docker image that contains all necessary python libraries like Tensorflow and the seismic modeling code SeisCL.

You first need to install the Docker Engine, following the instructions here. To use GPUs, you also need to install the Nvidia docker. For the later to work, Nvidia drivers should be installed. Then, when in the project repository, build the docker image as follows:

docker build -t seisai:v0

You can then launch any of the python scripts in this repo as follows:

docker run --gpus all -it\
           -v `pwd`:`pwd` -w `pwd` \
           --user $(id -u):$(id -g) \
           seisai:v0 Case_article.py --logdir=./Case_article

This makes accessible all gpus (--gpus all), mounting the current directory to a to the same path in the docker (second line), running the docker as the current user (for file permission), and runs the script Case_article.py.

b) Install all requirements

It is recommended to create a new virtual environment for this project with Python3. Install all requirements with

pip install .

The main Python requirements are:

• tensorflow
• SeisCL. Navigate to the directory of the SeisCL installation and follow the instruction in the README. Preferred compiling options for this project are api=opencl (use OpenCL, which is faster than CUDA for small models) and nompi=1, because no MPI parallelization is required. Be sure to install SeisCL's python wrapper.