Leveraging U-Net Variants for High-Precision Flood Mapping: A Performance Evaluation on the ETCI Dataset

Flood mapping is a crucial task in mitigating the adverse impacts of flooding by providing accurate spatial information about inundated areas. This study explores flood mapping using the ETCI dataset, leveraging the U-Net model and its enhanced variants. Initially, a basic U-Net architecture is employed to generate flood maps, followed by progressively improved versions incorporating batch normalization, dropout, and attention mechanisms. These modifications aim to enhance model generalization, robustness, and feature extraction capabilities. For the purpose of evaluating the model's performance, metrics like accuracy, F1-score, and Intersection over Union (IoU) are utilised. The generated flood maps demonstrate the potential of advanced U-Net architectures for precise and efficient flood delineation, providing a valuable tool for disaster management and planning.