Building dataset for SR

 \This is a comprehensive super-resolution building dataset, consisting of images sourced from three widely used building datasets: Wuhan University (WHU), Massachusetts (MAS), and Waterloo (WAT). Each of these datasets contributes unique characteristics, making this composite dataset more diverse and representative of real-world variations in building structures, layouts, and textures. The WHU dataset primarily focuses on high-resolution urban building imagery with well-annotated labels, providing a strong foundation for detailed feature extraction. The MAS dataset, originally designed for segmentation tasks, offers satellite images with varying resolutions and complex urban landscapes, making it a valuable addition for testing the generalization capability of SR models. The WAT dataset, known for its inclusion of diverse architectural styles and environmental conditions, further enhances the robustness of this dataset by introducing variations in lighting, shadowing, and image acquisition angles. By combining these three datasets, our work enables a more reliable and adaptable super-resolution training and evaluation process, ensuring that the proposed SR network can handle different architectural styles and urban environments more effectively.