Global Co-seismic InSAR Dataset for Deep Learning

A global InSAR image dataset for deep learning-based coseismic deformation detection. It is composed of 14,000 coseismic InSAR image patches (224 pixels × 224 pixels) obtained by preprocessing and manually annotating 343 raw coseismic interferograms (from Sentinel-1) of 62 medium-to-large earthquakes worldwide from 2015 to 2024. The primary purpose of this dataset is to support the development and training of deep learning models for coseismic deformation detection and analysis using InSAR data. It is a valuable resource for researchers in the fields of geophysics, earthquake engineering, remote sensing, and deep learning.