search

andreasMazur / geoconv

A Python library for end-to-end learning on surfaces. It implements pre-processing functions that include geodesic algorithms, neural network layers that operate on surfaces, visualization tools and benchmarking functionalities.
GNU General Public License v3.0
28 stars 2 forks source link

Pre-process dataset and pytorch example #4

Closed wq1999 closed 2 months ago

wq1999 commented 2 months ago

Hello, this work is very nice!! I only have point cloud data(N*3, N points with xyz coordinate), what should i do if i want to apply geoconv on my data for prediction task with pytorch?

andreasMazur commented 2 months ago

Hello! Thanks for trying out GeoConv. As of yet, GeoConv's preprocessing utility requires you to have triangle meshes. If you have pointcloud-data, you could try to first convert that data to triangle meshes using surface reconstruction techniques:

https://www.open3d.org/docs/latest/tutorial/Advanced/surface_reconstruction.html

Afterwards, check whether you have manifold-meshes:

https://www.open3d.org/docs/latest/python_api/open3d.geometry.TriangleMesh.html#open3d.geometry.TriangleMesh.get_non_manifold_edges

Otherwise you will not be able to compute GPC-systems properly.

Lastly, you should be able to preprocess your shapes as described by the provided examples in the geoconv_examples-folder.

Best of luck with your experiments!