Speeding up FPFH generation

[fradovitzky]

Hi,
I have successfully used FGR to align a pointcloud (~27000 points) with a transformed (rotated and translated) version of itself and it worked.

I noticed that the FPFH calculations take extremely long relative to the actual FGR alignment itself.

I am trying to use FGR for a real-time application (aligning consecutive pointclouds from a bag file), and clearly the speed of the feature detection will not do.

Any ideas on how I could speed this up? Is there any way I could use CUDA/GPU to accelerate this?

Would it be possible to "save" features found with FPFH for faster FPFH computation of a subsequent pointcloud (in a real-time application)?

[qianyizh]

We are transiting the code in this repo to a new repo called Open3D. There is a new FPFH computation and a new FGR.
http://www.open3d.org/docs/release/tutorial/Advanced/global_registration.html
http://www.open3d.org/docs/release/tutorial/Advanced/fast_global_registration.html

They should be faster than the implementation here - but probably not to an order of magnitude.

The Open3D team is working on a CUDA reimplementation of everything. Expect it to be available in a few months.