FloodNet: High-Resolution UAV Flood Scene Understanding
Advancing automated post-disaster damage assessment through semantic segmentation on high-resolution unmanned aerial vehicle (UAV) imagery from the benchmark FloodNet dataset.
Developing deep learning models to pixel-accurately classify flood-affected regions including inundated areas, flooded vs. non-flooded roads, buildings, vehicles, and vegetation.
Evaluating and optimizing state-of-the-art encoder-decoder architectures — U-Net, DeepLab, and SegNet — for aerial remote sensing constraints.
Translating aerial scene understanding into actionable spatial intelligence for AI-driven disaster management and rapid rescue routing.