MIRAGE: a geomorphology-constrained dataset of DInSAR wrapped phase signals from active mass movements

The MIRAGE dataset contains 4910 DInSAR wrapped phase displacement signals associated with active mass movements, extracted from 92 wrapped interferograms. This dataset has been specifically designed to train deep learning models on the automatic detection and classification of mass movements, following rigorous geomorphological criteria during signal interpretation. The analysed areas are located in the Central European Alps and the Northern Apennines (Italy). The dataset is organised into two folders, one per study area, each containing the following three products:1. Wrapped interferograms in TIF format.2. Coherence maps in TIF format corresponding to each interferogram.3. Labels in shapefile format, with one file per interferogram, containing the bounding boxes of displacement signals along with the seven descriptive attributes. The interferograms and coherence maps were generated using SARscape and GMTSAR for the Alps and Apennines areas, respectively. A more detailed description of the MIRAGE dataset is provided in the associated manuscript available as a preprint at https://doi.org/10.31223/X5877Z.