AI-Powered Landslide Region Mapping
This project uses deep learning and semantic segmentation to automatically map landslide regions from satellite and UAV imagery, supporting disaster management and geospatial analysis.
Highlights
- Semantic segmentation models trained and benchmarked on the CAS Landslide Dataset (UAV and satellite) and Landslide4Sense.
- Modular pipeline for data preprocessing, model training, evaluation, and K-fold cross-validation.
- Trained PyTorch models exported to ONNX for deployment.
- Interactive Streamlit app to upload images, run inference, and visualize segmentation results.
Tech stack: Python · PyTorch · Semantic Segmentation · ONNX · Streamlit · Remote Sensing