Simulated time-series InSAR data depicting coseismic deformation

Published: 31 March 2023| Version 1 | DOI: 10.17632/73h3d8sdhc.1
xue li,
chisheng wang,
chuanhua zhu,
Bochen Zhang,
baogang li


This dataset contains the input, output data, and related codes used to train and test the autoencoder-based deep learning model for extracting coseismic deformation from InSAR interferograms. The dataset includes a total of 28800 samples, each consisting of 9 two-dimensional interferometric synthetic aperture radar (InSAR) images. The images were generated using Matlab software and simulated the characteristics of InSAR interferograms, including the effects of atmospheric and topographic interference. The dataset is organized into two folders: 'Training&Testing' and 'Prediction'. The 'Training&Testing' contains three training sets of 14400,9600,4800 samples and corresponding testing sets. The 'Prediction' contains real-world data for model validation. Each folder includes the corresponding input, output data, and codes. The input data are the time-series interferograms, which are saved as .mat files. The dataset also includes a README file with detailed information on how to use the data and reproduce the experiments. This dataset can be used to train and evaluate the deep learning models for coseismic deformation extraction from InSAR interferograms. The dataset is made publicly available on Mendeley Data and can be accessed at 'DOI: 10.17632/73h3d8sdhc.1'.



China University of Petroleum Huadong, Shenzhen University


Synthetic Aperture Radar Images, Satellite Geodesy, Earthquake Hazard