How to speed up TriangleMesh::CreateFromPointCloudBallPivoting function?

[Cassius0924]

Checklist

• [X] I have searched for similar issues.
• [X] For Python issues, I have tested with the latest development wheel.
• [X] I have checked the release documentation and the latest documentation (for master branch).

My Question

I'm working on a project that processes large point clouds (over 1 million points) and needs to frequently create triangle meshes. The current TriangleMesh::CreateFromPointCloudBallPivoting implementation is too slow for my use case.

I noticed that many other functions in Open3D have tensor-based implementations (like ICP registration) that utilize GPU to accelerate computation. It would be great if we could have a GPU-accelerated version of ball pivoting using CUDA or tensors.

[theNded]

Unfortunately, this approach is not designed for parallel processing, as the operations on a point will directly affect its neighbors, leading to race conditions. In comparison, Poisson reconstruction is relatively easier to parallelize. For instance, there is such a repo (https://github.com/DavidXu-JJ/PoissonRecon_GPU). At current we do not have bandwidth to incorporate it, so you may want to make use of the repo indepedently as an option. Hope it helps!