GET STARTED

PoreSeg AI

Pore Space Segmentation Detection Model in Rocks

We developed a deep learning solution for segmenting and analyzing pore spaces within rock samples for the petroleum industry. Using 3D imaging techniques, the system quantifies porosity and permeability critical parameters for evaluating oil and gas reservoirs enabling faster, more consistent analysis of complex rock structures.

What We Built

A comprehensive computational geoscience pipeline using:

  • 3D imaging processing for volumetric rock sample analysis
  • Deep learning segmentation models for pore space identification
  • U-Net architecture with 3D inception blocks for volumetric segmentation
  • Data augmentation techniques for improved model generalization
  • Cloud-based infrastructure for GPU-accelerated processing

The system acts as an automated geological analysis engine, dynamically segmenting pore spaces and quantifying reservoir properties from 3D imaging data.

Key Features​

3D Image Segmentation

  • Deep learning-based pore space identification
  • Volumetric analysis of rock samples
  • High accuracy in complex pore structures

Rock Flow Properties

  • Automated calculation of porosity percentages
  • Permeability estimation from pore structure
  • Critical parameters for reservoir evaluation

Generalization Across Rock Types

  • Training on diverse rock samples
  • Adaptive processing for different lithologies
  • Robust performance across varying microstructures

Scalable Processing

  • GPU-accelerated segmentation
  • Cloud infrastructure for large datasets
  • Efficient handling of 3D volumetric data

How It Works

3D imaging data of rock samples is ingested

Data is preprocessed and normalized for model input

Deep learning segmentation model identifies pore spaces

Volumetric analysis calculates porosity percentages

Pore connectivity analysis estimates permeability

Results are compiled for reservoir evaluation and decision-making

Results

Automated

pore space segmentation with high accuracy

Consistent

analysis across diverse rock types

Reduced

time for geological characterization

Improved

reservoir evaluation and resource extraction planning

Value Dilevered

This project transformed manual geological analysis into an automated, AI-powered process. By applying deep learning to 3D rock imaging, the solution enables faster, more consistent characterization of pore structures supporting improved decision-making in petroleum exploration and production, optimizing drilling techniques, and enhancing reservoir management efficiency.

Want to build AI-powered Problem solutions?

View Our Work

#Deep Learning #Computational Geoscience #Pore Space Segmentation #3D Imaging #Petroleum Engineering #U-Net #Reservoir Characterization #Geoscience AI