Self-Supervised Annotation of Seismic Images using Latent Space Factorization

TL;DR

This paper introduces a self-supervised framework that automates seismic image annotation by factorizing the latent space of a deep network, enabling delineation of geological structures with minimal supervision.

Contribution

It presents a novel latent space factorization method for seismic image annotation using only image-level labels, reducing reliance on manual pixel annotations.

Findings

Effective delineation of geological elements demonstrated

Qualitative comparison shows improved annotation quality

Framework reduces annotation effort and subjectivity

Abstract

Annotating seismic data is expensive, laborious and subjective due to the number of years required for seismic interpreters to attain proficiency in interpretation. In this paper, we develop a framework to automate annotating pixels of a seismic image to delineate geological structural elements given image-level labels assigned to each image. Our framework factorizes the latent space of a deep encoder-decoder network by projecting the latent space to learned sub-spaces. Using constraints in the pixel space, the seismic image is further factorized to reveal confidence values on pixels associated with the geological element of interest. Details of the annotated image are provided for analysis and qualitative comparison is made with similar frameworks.