Volumetric Supervised Contrastive Learning for Seismic Semantic Segmentation

Dataset

The dataset is the labeled F3 Block of the Netherlands from: https://arxiv.org/pdf/1901.07659.pdf

Code Usage

1. Initiate contrastive learning pre-training with main_supcon.py. Identify the labels to train with by choosing the appropriate discretization level csv file.
2. Do semantic segmentation training within main_seismic_semantic.py by identifying the checkpoint file generated in step 1 and choosing the test fold csv file of interest.

Links

Associated Website: https://ghassanalregib.info/

Code Acknowledgement: Code is based off of https://github.com/HobbitLong/SupContrast.git.

Citations

If you find the work useful, please include the following citation in your work:

@inproceedings{kokilepersaud2022volumetric,
title={Volumetric Supervised Contrastive Learning for Seismic Semantic Segmentation},
author={Kokilepersaud, Kiran, and Prabhushankar, Mohit and AlRegib, Ghassan},
booktitle={The International Meeting for Applied Geoscience & Energy},
year={2022}
}