Program Overview
Seismic interpretation inherently involves integrating diverse views and modalities such as multiple directions of subsurface images, well-log data, and diverse attributes, to delineate subsurface structures and estimate reservoir properties. Sparse labels from costly drilled wells further complicate interpretation. This research program addresses these key challenges through advances in multi-view and multi-modal learning to deliver dense and high-quality interpretations of large seismic volumes.
Objectives
- Cross-Modal Representation Learning: Develop techniques adapted from multi-modal video understanding and medical applications to the seismic domain. Integrate and align disparate data modalities into shared representations.
- Sigmoid with Implicit Background Estimation (SCrIBE)-based Models: Aims to project each modality into a shared latent space by aligning other modalities with the 3D seismic model modality.
- ImageBind-based Models: Aims to bind all seismic modalities into a unified latent space without the need for paired datasets.
- Federated Learning Frameworks: Develop decentralized training frameworks in the presence of data heterogeneity of multi-view data from geographically dispersed regions or diverse partitions of the same volume.
- Develop 3D Multi-View Fusion Frameworks: Develop models that learn to fuse multiple and complementary views of seismic data to a unified volumetric representation.
- Leverage Sparse Well Data via Weak Supervision: Create strategies that utilize sparse well-log data and seismic volumes through consistency constraints across modalities.
Key themes
- Multi-modal learning.
- Multi-view representation learning.
- Cross-modal representation learning.
- Federated learning.
Participation and Governance
This CRP operates within the ML4Seismic philanthropic partnership framework. Participation requires enrollment as an Executive Member with at least one CRP selected. Partners are
invited to:
- Participate in CRP-specific workshops and benchmarking initiatives
- Collaborate on shared scientific objectives and data challenges
- Engage with students and researchers in community events
All research outcomes are shared via open-source repositories and peer-reviewed publications in accordance with ML4Seismic operating guidelines
Contact
Prof. Ghassan AlRegib alregib@gatech.edu
