Natural disasters are extreme and sudden events that pose significant risks to human life and infrastructure. In the context of natural disaster management, the ability to rapidly map the affected areas is essential for risk assessment. This thesis investigates the use of deep learning techniques for landslide detection using InSAR data. The study is structured into two main parts. The first part focuses on the selection of landslide events and the construction of the corresponding dataset, following a processing pipeline that generates InSAR products from SAR data acquired by the Sentinel-1 mission. The dataset includes the following products: phase, coherence, intensity, displacement, unwrapped phase, elevation, and local incidence angle, combined with landslide ground truth and validity masks to account for missing or non-informative regions.