Real-Time Flood Water Segmentation with Deep Neural Networks

Nowadays, extreme floods pose severe threats to human lives and infrastructure, necessitating effective flood disaster management plans and systems. In this paper, we address the crucial need of real-time flood monitoring by leveraging computer vision models. To this end, benchmark deep neural networks are trained on the flood water segmentation task, using a novel dataset that was created by combining and annotating flood images from different sources. Our experimental evaluation of two real-world flood videos showcases the potential of computer vision in fast and accurate flood monitoring. Moreover, we investigated the use of semi-supervised training methods to enhance the flood segmentation performance by taking advantage of large unlabeled datasets. Our work emphasizes the potential of applying state-of-the-art big visual data analytics tools to mitigate the devastating impacts of floods on communities worldwide.