Landslide Detection in Satellite Images using InceptionU-Net and Convolutional Block Attention Module

In the contemporary world, Landslides pose huge risks to communities and infrastructure globally. One of the reasons it occurs is due to slope movement caused by urbanization and climate change in recent years. Landslide Detection Systems (LDS) are an evolving technology that is very crucial for handling the impacts of these natural disasters. It is essential for detecting and monitoring landslide occurrences across diverse terrains. This paper aims to propose one such automated Landslide Detection System based on the novel idea of integrating InceptionU-Net and Convolutional Block attention module that can efficiently detect landslides in satellite imagery using a benchmark dataset called LandSlide4Sense. The motive is to determine the most effective methods by training them on huge amounts of data which enhances accuracy and ensure quick detection and response. Deep Learning techniques like U-Net, ResUNet, U-Net++, InceptionU-Net, Inception, DenseUNet, and U2Net are utilized and an additional contribution is made using attention mechanisms on these models. After testing these models on the test data and comparing the models using metrices such as Loss, Accuracy, F-score, IOU, Precision, and Recall. With a particular emphasis on F-score and IOU, it was found that the Inception-UNet model integrated with attention mechanisms proves to perform the best with the highest Fscore of 72.8% and IOU of around 60%. Through this research, the advancements in DL-based semantic segmentation for satellite imagery analysis are highlighted.