SeisSegDiff

An efficient few-shot segmentation diffusion model for seismic facies classification

This code is based on Label-Efficient Semantic Segmentation with Diffusion Models and guided-diffusion.

Note: use --recurse-submodules when clone.

 

Publication

This work has resulted in 2 published papers in Computers & Geosciences: SeisSegDiff: A label-efficient few-shot texture segmentation diffusion model for seismic facies classification and Evaluating key parameters impacting the performance of SeisSegDiff model for seismic facies classification.

 

Overview

The work investigates the use of diffusion models to enhance the generalization capabilities and accuracy of deep learning models for seismic facies segmentation.

 

Datasets

The evaluation is performed on 2 datasets: Pari and F3. These are popular open-source seismic datasets from New Zealand and the Netherlands respectively. The Parihaka data from New Zealand: (a) 3D seismic data, (b) facies labeled by expert interpreters, and (c) the percentage of individual facies in the data.

The F3 data from the Netherlands: (a) 3D seismic data, (b) facies labeled by expert interpreters, and (c) the percentage of individual facies in the data.

 

DDPM

Pretrained DDPMs

The diffusion model trained on both Pari and F3 can be downloaded from Tobi_model.

Run

1. Download the datasets:
     bash datasets/download_datasets.sh
2. Download the DDPM checkpoint:
      bash checkpoints/ddpm/download_checkpoint.sh <checkpoint_name>
3. Check paths in experiments/<dataset_name>/ddpm.json
4. Run: bash scripts/ddpm/train_interpreter.sh <dataset_name>