Can AI Learn from Its Own Interpretations? Transfer Learning and Human-in-the-Loop Seismic Interpretation in the URGENT project One of the challenges in applying Machine Learnin...
Can AI Learn from Its Own Interpretations?
Transfer Learning and Human-in-the-Loop Seismic Interpretation in the URGENT project
One of the challenges in applying Machine Learning to seismic interpretation is that geological settings vary significantly from basin to basin. Models trained in one area often require adaptation before they can be used effectively elsewhere.
As part of the Horizon Europe funded URGENT consortium, we have been developing an integrated AI-assisted seismic interpretation workflow within OpendTect that combines:
• Transfer-trained 3D machine learning models for Fault Likelihood, Fault Dip and Fault Strike prediction • Automated fault surface reconstruction using FaultStickTracker • Interactive model retraining directly from interpreter feedback • Transformer-based Seismic Foundation Models trained on ~110 seismic surveys
The workflow creates a practical human-in-the-loop interpretation cycle. Initial fault predictions are converted into fault surfaces, interpreters perform QC and refinement, and the resulting interpretations are automatically reused to improve the prediction model. This allows models to adapt to new geological settings while maintaining interpreter control and geological consistency.
In parallel, the Seismic Foundation Model uses self-supervised transformer architectures to learn generalized structural and stratigraphic representations from approximately 1.9 million seismic training samples. The objective is to improve transferability across surveys and provide a foundation for future seismic interpretation tasks such as fault detection and facies classification. The goal is not to replace interpreters, but to reduce repetitive manual work and allow geoscientists to focus on geological decision making rather than data generation.
dGB Earth Sciences is proud to contribute these developments to the URGENT project and the broader effort to accelerate geothermal exploration in urban environments through advanced subsurface technologies.
#OpendTect #AI #seismic #ML #URGENTProject #Innovation #Technology #Geophysics #Geosciences #GeothermalExploration #GeothermalEnergy #HorizonEurope #HEU #CleanHeating #EUProjects #Decarbonisation #RenewableEnergy #Sustainability #EuropeanCommission #EuropeanUnion #EU
