Pixels Rock: A Journey from Simple Segmentation to Deep Learning Techniques to Unravel Deformation in Rocks
Short Abstract: In the presentation "Pixels Rock: a journey from simple Segmentation to Deep-Learning Techniques to unravel deformation in rocks," we embark on a comprehensive exploration of how we can use image analysis methodologies to study deformation and failure localisation in rocks. We will start presenting 2D image analysis and how we can use it for fracture segmentation, contrasting “static” analysis techniques with the “dynamic” nuances offered by 2D Wavelets. Progressing into the realm of 3D and 4D image analyses, the talk shows in-situ and operando imaging techniques that capture the evolution of damage over time. A significant highlight is the introduction of Deep Learning for 3D Image Segmentation, offering a transformative approach to understanding how stress influences metamorphic processes. With a focus on permeability modelling, we emphasize its pivotal role in geophysics and structural geology. Concluding the talk, a glimpse into the future perspectives sketches the horizon of advancements and potential applications in the field.