What is it about?
Detecting the onset of seismic signals is crucial, but it involves significant human labor. We developed a self-trained convolutional neural network (CNN) model to detect the onset of seismic signals automatically. In experiments on an open seismic dataset, our picking method outperforms the benchmark on most evaluation metrics.
Featured Image
Photo by Marco Reyes on Unsplash
Why is it important?
Our detection model effectively learns across various domains, filters out incorrect predictions, and assigns pseudo first-arrival times to unlabeled data during training. Owing to these features, our approach requires only a minimal amount of labeled data and remains robust even when applied to data from diverse regions. As a result, our method contributes to reducing human labor and the total seismic processing time.
Perspectives
I believe that self-learning in different domains can be applied not only to seismology but also to other fields. I hope you find this article thought-provoking.
Mitsuyuki Ozawa
JGI, Inc.
Read the Original
This page is a summary of: Automated picking of seismic first arrivals using a single- to multidomain self-trained network, Geophysics, October 2023, Society of Exploration Geophysicists,
DOI: 10.1190/geo2022-0666.1.
You can read the full text:
Contributors
The following have contributed to this page
