Road extraction from remote sensing images is a specific application of the broader concept of road extraction, where the goal is to identify and delineate road networks from images captured by remote sensing devices such as satellites. It is a difficult task due to the diversity of road looks in different conditions and the existence of other objects that can be mistaken for roads. To fix the issue, we proposed applying a state-of-the-art deep learning approach for road extraction from remote sensing images to detect the roads in various environments because roads are necessary for transportation, infrastructure development, and emergency response. Accurate road maps are required for navigation, routing, and traffic management. Our proposed approach is evaluated on a dataset of remote sensing images from different locations and different conditions, So, in this system, we have expertise in image processing and deep learning frameworks. In this research, we apply one hot encoding mask to preprocess the dataset. Subsequently, we have successfully implemented pre-trained models such as FCN, PSPNet, U-Net, SwinU-Net, LinkNet, and DeepLabV3 models. Notably, the SwinU-Net achieved the highest F1-score 0.95%.