niessner / Matterport

Matterport3D is a pretty awesome dataset for RGB-D machine learning tasks :)
https://niessner.github.io/Matterport/
MIT License
951 stars 154 forks source link

point cloud of the building #8

Open hope-yao opened 6 years ago

hope-yao commented 6 years ago

Hi Matterport3D team,

Thank you for releasing this great dataset! I'm wondering if there is a point cloud representation of the building available? Or are there any easy ways you suggest that I can do to convert the depth image/ mesh into (high quality) point clouds? Thanks!

DBobkov commented 6 years ago

A proper way to convert a mesh into point cloud representation is to use mesh_sampling (https://github.com/PointCloudLibrary/pcl/blob/master/tools/mesh_sampling.cpp), which will generate points uniformly from the faces of the mesh.