Building Polygon

Building polygon extraction from remotely sensed imagery focuses on automatically generating precise vector representations of buildings, crucial for high-resolution mapping and 3D modeling. Current research emphasizes end-to-end deep learning approaches, employing architectures like R-CNNs and graph neural networks to directly predict building polygons or their constituent vertices and corners, often incorporating techniques like contrastive learning to improve accuracy and efficiency. These advancements improve the speed and accuracy of building database updates, enabling more efficient urban planning, infrastructure management, and other applications requiring detailed building information.