A novel convolutional neural network model with hybrid attentional atrous convolution module for detecting the areas affected by the flood

Natural disasters are sudden and unexpected events that occur as a result of natural processes and often have a significant impact on people, animals, and plants, often resulting in material and emotional losses. Two of the most devastating disasters are floods and tsunamis. After these disasters, search and rescue operations to identify flooded areas are very important to save the lives of affected people and to ensure the health and safety of rescue workers. In this study, a novel semantic segmentation model called Flood Area Segmentation Network (FASegNet) was proposed to guide search and rescue teams in disaster areas to speed up search and rescue operations after natural disasters such as floods, high tides, and tsunamis. The obtained results were compared with common image segmentation models used in various fields. It was found that the developed model achieved higher accuracy rates with fewer parameters. When tested with the "Flood Area" and "Water Body" datasets without pretraining and data augmentation, FASegNet achieved mIoU accuracy of 84.3% and 84.5% respectively.