Diffusion models in Porous Media (Diff-Twins)

Modelos probabilísticos de difusão e algoritmos multifidelidade para o desenvolvimento de gêmeos digitais de estruturas porosas e sistemas oceanográficos

Unlock the future of porous media analysis with a groundbreaking research project led by L'Cadame in partnership with ExxonMobil. The central challenge facing our field today is the accurate reconstruction of 3D porous media images from low-resolution 2D or 3D samples that may contain defects. Such inaccuracies can drastically compromise the quality of research and lead to incorrect operational decisions, affecting a variety of critical operations from resource extraction to environmental assessments. Our innovative approach employs advanced diffusion probabilistic models and score-based models to enhance inpainting, super-resolution, and uncropping methods. Taking it a step further, we're integrating these models with specialized machine learning techniques dedicated to volumetric reconstruction. The end goal is a robust and precise 3D reconstruction of porous media that will significantly elevate the quality of data analysis and decision-making processes across various applications. Whether you're an industry specialist or an academic researcher, this project presents a unique opportunity to be at the forefront of a scientific breakthrough. Join us in pioneering the future of porous media analysis.