Generative Modeling of Disagreements for Expertise-Based Seismic Fault Labels
In this work, we use Amazon Mechanical Turk to elicit various labeling patterns between experts and practitioners, and propose a generative modeling approach to characterize such expert-practitioner label discrepancies.
Interpretational Uncertainty-Based Seismic Fault Labeling With Reduced Expert Annotations
This work leverages crowd-sourced practitioners to enhance seismic fault interpretation, reducing the requirement for costly expert annotations.
Improving Seismic Interpretation Accuracy and Efficiency With Human-Machine Collaboration
This work introduces a new framework on adding prompts to segment salt dome region using the Segment Anything model (SAM)
Multi-View Seismic Segmentation With Test-Time Augmentation
This work introduces a method that integrates dynamic uncertainty thresholding with an encoder-decoder CNN to refine seismic facies segmentation predictions across multiple seismic views, inline, crosslines, and depth, enhancing segmentation accuracy.
Uncertainty-Aware Seismic Fault Annotations Can Reveal Insights Into Label Expertise and Confidence
We present a seismic fault segmentation dataset compiled across 3 different levels of confidence and expertise. Each image was labeled by 38 annotators: 29 novices (N), 8 practitioners (P) and 1 expert (E), leading to over 15000 image annotations. The acquired data from each annotator includes the detected fault regions, the associated confidence level, and the time taken for the labeling of each image.
Optimizing Prompting for Foundation Models in Seismic Image Segmentation
In this work, we aim to find this optimal prompting solution for a task by measuring the level of importance of each prompting point, coming up with an overall objective rating of the entire strategy based on the concept of sufficiency.
Integrating Granularity Information Into Seismic Contrastive Learning
In this work, we demonstrate that appropriately integrating granularity information into volumetric representation learning can improve seismic representations for downstream interpretation tasks.
Assessing Seismic Data Through Self-Supervised Representational Analysis
In this work, we identify representational-based metrics that can characterize why certain seismic sections perform better or worse on downstream segmentation tasks without relying on access to label information.