Image segmentation is crucial for flood analysis because it helps identify flood-prone regions that require attention from NGOs and government aid. It becomes challenging to scout small and narrow flood-affected areas, particularly in countries like India with lush forests and limited visibility, especially in low-light settings, surrounded by borders on three sides. In this comparative study, we evaluate the performance of two prominent deep learning architectures, UNet and U2Net, in the context of flood image segmentation. Leveraging a diverse of flood images, our analysis assesses their effectiveness in accurately identifying and delineating flooded areas in aerial imagery. While UNet demonstrates robustness in scenarios with well-defined flood boundaries, U2Net excels in capturing subtle flood patterns amidst complex backgrounds. The study’s findings provide valuable insights for selecting the most suitable model based on specific flood monitoring needs, ultimately enhancing flood management and disaster response efforts for flood-prone regions.