Mamba for Landslide Detection: A Lightweight Model for Mapping Landslides With Very High-Resolution Images
Citation
Tang, L., Lu, W. (2024). Mamba for Landslide Detection: A Lightweight Model for Mapping Landslides With Very High-Resolution Images. IEEE Geoscience and Remote Sensing Letters. Link to paper
Abstract
A lightweight landslide detection method based on the newly proposed Mamba network with an encoder-decoder structure achieves superior landslide detection accuracy, with approximately 2% improvement in F1 score across various scenarios over conventional models, while significantly reducing computational costs. Mamba, based on the state space model, has drawn considerable attention due to its ability to achieve linear complexity while maintaining high modeling precision. CNN models focus primarily on local features, often missing crucial global context in landslide images. Conversely, Transformer-based models excel at capturing global features but are hindered by high computational complexity. The research indicates Mamba-based models offer a promising solution for efficient, accurate landslide detection in remote sensing applications.