Figure 1: Given a 3D model (left), our network creates a line drawing that conveys its structure more accurately compared to using other methods individually, such as Occluding Contours [1], Apparent Ridges [22], and pix2pix variants [37]. Please see appendix for results on many more models.
We introduce a dataset of 3D building meshes with annotated exteriors (top). We also present a graph neural network that processes building meshes and labels them by encoding structural and spatial relations between mesh components (bottom). Our dataset also includes a point cloud track (blue box). Examples of erroneous network outputs are in red text.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.