Potholes with stagnant water are one of the major causes of accidents and vehicle damage or tire damage, and they also contribute to an increase in harmful diseases such as dengue and malaria by nurturing the growth of carriers such as mosquitoes. Thus, there is a need to effectively monitor the road conditions and repair them. Though this is a very manual task, technology can offer assistance. In this paper, we have proposed a semantic segmentation technique to identify stagnant water and potholes filled with stagnant water using a modified version of DoubleU-Net. The results showed the Mean IOU as 86% and it needs 0.18 s to detect stagnant water in each image. This approach delivers a satisfactory result in detecting stagnant water and thus can be developed and implemented as a module in automobiles and road assessment systems.