Computer Vision with Deep Learning for Subsurface Characterisation
Carbonate rocks host over half of the world's reservoirs and are critical for clean energy transition technologies like carbon capture and storage, geothermal energy, and hydrogen storage, yet their characterization remains qualitative, time-intensive, and subjective. In this talk, I present a series of studies from our research group applying deep learning to automate and quantify carbonate core analysis, with four objectives: identifying Dunham textures using convolutional neural networks while expanding and balancing training datasets, detecting and quantifying individual carbonate grain types through object detection models, reducing reliance on labelled data via transfer learning and semi-supervised learning, and leveraging generative AI (Pix2PixHD) to translate Formation MicroScanner image logs into synthetic core images for more accessible subsurface interpretation. Collectively, these studies demonstrate both the promise and pitfalls of deep learning for carbonate sediments, with strong potential to transform core analysis across academic research and industrial applications in CCS, resource exploration, and paleoenvironmental reconstruction.
Short Bio: Professor Cédric John is Head of Data Science and the Environment at the Digital Environments Research Institute (DERI) at Queen Mary University of London, having previously held a faculty position at Imperial College London. He holds a PhD in Geology and his research bridges carbonate sedimentology, clumped isotope geochemistry, field geology, and data science, including artificial intelligence. His work addresses two interconnected challenges: understanding past and present climate change and leveraging the subsurface to accelerate the transition to cleaner energy. More information about his team's work is available at https://www.john-lab.org/
To get news, invitations to seminars and more from Njord, please go here to subscribe to our newsletter.