This research addresses the problem of automatic extraction of topographical elements including roads, water bodies, mountains, and ground surfaces from satellite images with the application of deep learning algorithms. Such terrains are very important in urban planning as well as in environmental management and control of disasters. So, the final outcome of the project is aimed at creating a viable model that will help to detect and classify these kinds of landscape objects to minimize human work and increase precision. To accomplish such tasks, we applied two state-of-the-art deep learning networks, U-Net and Deep Lab v3, for feature extraction. The model will be trained on Deep Globe 2018 dataset, a Sentinel imagery dataset, both easily obtainable in high-resolution satellite images. Some preprocessing operations, such as image segmentation, were carried out to boost the efficiency.