Underground Garage Point Cloud Dataset (UGD)

This dataset contains labeled 3D point clouds collected from two underground garage environments, including both under-construction and completed scenes. Annotated classes include walls, ceilings, floors, and MEP components, among others. The dataset is designed to support research in semantic segmentation, 3D reconstruction, and digital twinning. It is associated with the paper titled “A point cloud dataset and deep learning method for semantic segmentation of underground garages” accepted in [Computer-Aided Civil and Infrastructure Engineering].