Closed samanemami closed 2 years ago
As far as I understand, we have to build oc_tree after calling the class and the only method that works fine for this matter is build_octree(point)
Cause in any case we have to feed the octree with the points.
But there is a conflict between the size when applying the voxel with built octree
oc_tree = ocnn.octree.Octree(depth=8)
oc_tree.build_octree(point)
ocnn.nn.octree2voxel(tensor_cloud, oc_tree, 8)
RuntimeError: shape mismatch: value tensor of shape [841348, 3] cannot be broadcast to indexing result of shape [109320, 3]
please refer to the example files in this folder https://github.com/octree-nn/ocnn-pytorch/blob/master/projects/datasets to figure out the usage of build_octree
Thanks, But I could not find related examples to build_octree
, they used the Point class only.
I think we can relate this issue to #5
Please first run the provided examples following the tutorials like Classification, Segmentation
When I try to build an OCTree, it returns the following error.
The oc_tree has the following type