ILCD-India Landslide Change Detection Dataset for deep learning applications in landslide detection

Manuscript under preparation

Abstract
ILCD contains more than 1000 landslides mapped from sentinel 2 images using a semi-automated framework ML-CASCADE. To develop a fully automated landslide identification model we develop and compare multiple AI-based frameworks like ensembles(Random Forests), deep learning segmentation (SegNet, UNet), and change detection (TINY-CD). Since the ILCD images are of different shapes, for deep learning we developed a patching framework that can process images of any size for training and deployment Github codebase