Deep Neural Networks for Extraction of Buildings from Satellite Images

Building extraction from high-resolution satellite images is a crucial task for urban planning, disaster management, and land-use monitoring. This study uses deep learning techniques for accurate segmentation and localization of buildings using network architectures such as U-Net with MobileNet, and Transfer learning models including YOLOv8. This paper analyzes performance accuracy, precision and scalability of these deep learning models. This study shows how advanced deep learning methods may be used to extract buildings from satellite photos with accuracy and scalability. The aim if this paper is to provide a scalable and reliable method for accurately extracting buildings so that satellite images may be analysed effectively for a variety of real-world uses.