Nishan Bhattarai

Global flood extent mapping from Sentinel-1 SAR imagery using a UNet ++ deep learning framework with permanent-water-augmented input: evaluation across geographically diverse flood events

Sentinel-1 SAR and JRC water inputs feed a UNet++/EfficientNet-B4 model, producing a 3-class flood map (background, permanent water, flood) for a river bend, applied to disaster response; macro IoU 0.734, F1 0.829. Accurate and timely flood mapping is essential for disaster response, humanitarian coordination, and i...