Image Segmentation for Disaster Resilience is a deep learning project developed for the FloodNet Challenge, focused on leveraging semantic segmentation to assist in flood impact analysis. Using a U-Net architecture, the model segments aerial imagery to detect key features such as flooded buildings, roads, water bodies, vegetation, and more. The project empowers emergency response teams and urban planners by enabling fast, visual interpretation of flood-affected areas.
This repository includes model training code, preprocessing scripts, visualization tools, and a Streamlit-based demo app for real-time interaction.
** Key Features ** Multi-Class Image Segmentation: Differentiates between flooded and non-flooded structures and natural elements. U-Net Architecture: Optimized for pixel-level segmentation tasks. FloodNet Dataset: High-resolution imagery for robust model training and evaluation. Interactive Interface: Built with Streamlit for easy model testing and visualization.