Building a 3D mesh benchmark dataset 2019

The project aims on the collecting ground truth data for 3D semantic mesh. The idea is classify each triangle of a 3D textured meshes obtained from multi-view stereo or SfM into different classes: (roof surface, roof superstructure, facade surface, facade installment, trees, lower vegetation, ground, water, boat, vehicles, bridges, unknows.). For this purpose a (semi-)manually labelling tool from TU Delft will be used.