event

Latent Space Exploration of Neural Networks for Annotation Efficient Subsurface Interpretation and Characterization

Primary tabs

Title:  Latent Space Exploration of Neural Networks for Annotation Efficient Subsurface Interpretation and Characterization

Committee: 

Dr. AlRegib, Advisor    

Dr. Anderson, Chair

Dr. Davenport

Abstract: The objective of the proposed research is to explore the use of deep autoencoders and learn data manifolds to rank and characterize unlabeled training examples by their outof-distribution-score. It is theorized that this ranking can be used to select and annotate only the most informative training samples for seismic interpretation tasks, greatly reducing the annotation effort for seismic interpreters.

Status

  • Workflow Status:Published
  • Created By:Daniela Staiculescu
  • Created:04/09/2022
  • Modified By:Daniela Staiculescu
  • Modified:04/09/2022

Categories

Target Audience