The dataset contains 2.904 geometries of single-family houses in the form of annotated Point Clouds, and was developed in order to train 3D Generative Adversarial Networks. The geometries are segmented within 3 classes: wall, roof, floor. The points of the point clouds are saved in .pts files while their labels are saved in .seg files. The creation of the dataset was done in a semi-automated way that consists of two stages: a) creation of module geometries representing building components (done in Rhino3D) b) the conversion of the geometries into Point Clouds with the Cockroach plug-in. 25 wall modules and 35 roof modules were created. Each wall module was combined with each roof module. Data augmentation methods were applied to maximize the size of the dataset: the modules were scaled in 3 ranges, and rotated 90 degrees for a wider feature space. The dataset can be used to train 3D GANs with architecturally relevant data. Connected publication describing a use case of this dataset to follow.
Dato for tilgængelighed9 nov. 2022
ForlagMendeley Data