What is it about?
Training image is a key input of multi-point geostatistical modeling. For modeling sedimentary facies under non-stationary conditions, it is common to first generate non-stationary training images, then use a partitioned simulation approach, and finally merge the realizations of each sub-region. We propose a new method for partitioning non-stationary training images based on features extracted using CNN mode l(ResNet and DenseNet).
Photo by Dan Roizer on Unsplash
Why is it important?
It can reproduce well the spatial variation characteristics of non-stationary training images, and provides a new method for processing multi-point geostatistical non-stationarity.
Read the Original
This page is a summary of: Nonstationary training image partition algorithm based on deep features, Interpretation, September 2022, Society of Exploration Geophysicists, DOI: 10.1190/int-2022-0023.1.
You can read the full text:
The following have contributed to this page