Enhancing Building Extraction from Remote Sensing Images Using YOLOV8

Building extraction from satellite remote sensing images is a critical task with various applications. Despite recent advancements in convolutional neural networks (CNNs), challenges persist in accurately delineating building outlines and separating distinct buildings. In this study, We suggest employing the YOLO model, a widely recognized object detection system renowned for its speed and precision, for the task of extracting building structures. Our objective is to enhance both precision and efficiency by customizing YOLO for the endeavor of building segmentation. This customization is expected to optimize the model's performance in this particular field. We demonstrate its effectiveness in accurately identifying building footprints and separating individual buildings by comparing it with the most commonly used instance segmentation model, Mask R-CNN. Our results underscore the effectiveness of YOLO as a sturdy tool for extracting buildings from remote sensing imagery, presenting encouraging prospects for a range of applications such as urban planning and 3D reconstruction.