UNet

[NoOneUST]

Hello,

If I use SubMConv to extract features from a sparse point cloud, how should I recover the point cloud using a decoder? I only want to get values of the points existed in the original sparse point cloud.

[FindDefinition]

You need to keep point-to-voxel indices when do point2voxel, then you can use it to recover point cloud. You can use generate_voxel_with_id from PointToVoxel to get that tensor.