Unmanned aerial vehicles (UAVs) have become a crucial tool in daily life, especially when it comes to managing damage caused by natural disasters such as earthquakes and floods. The optimal choice of UAVs in natural disasters due to its effective role in identifying the affected areas for fast damage management. Semantic segmentation of UAVs images is a significant challenge in the computer vision field; to overcome this issue many deep learning models were involved in the study, in particular models based on convolutional neural network (CNN). Recently transformer-based techniques demonstrated its effectiveness in semantic segmentation tasks. In this work, we aimed to use semantic segmentation for post-disaster images taken by unmanned aerial vehicles (UAVs) and evaluated the performance of vision transformers models designed for semantic segmentation on the FloodNet dataset, which contains high-resolution images taken from low altitudes for flooded and non-flooded areas and compare their performance to Unet, DeeplabV3 + and FCN. For this purpose, we used SegFormer series models (SegFormer-B1, SegFormer-B2 and SegFormer-B3) with transfer learning technique for better performance. Effectively reducing the impact of flooding, our study can be used to develop disaster response systems.