Wrong aligment on full room scans

[tamireiny]

Hello,
I would like to publish a comparison between FGR and my 3D alignment method.
I conducted an experiment of 12 tests, each test contains scene + template, each 3D Model can be represented as polygon mesh or point cloud.
10 of the scenes are from sceneNN and 2 from scanNet, the templates are taken from shapeNet.
The alignment presented is between the scene (source) and template (target) point clouds.
I used the voxel radius of 0.04 and the 1:2:5 radios are suggested in the troubleshoot.
The parameters of FGR remind the defaults.

In the link below I have visualized the transformation T generate by FGR for all 12 tests. It seems that in all tests the result is not successful. Can you suggest general parametrization that will yield a much better alignment for the tests?

In the folder you can find the python script I used to create the FPFH and run FGR.

All data is here: https://www.dropbox.com/sh/dxglr4ga1f2ly0k/AACIE4eNcCYBMhYgy3qjbO_7a?dl=0

Thanks,
Tamir

[syncle]

Hi Tami

I conducted an experiment of 12 tests, each test contains scene + template, each 3D Model can be represented as polygon mesh or point cloud.

I think it is because of FPFH feature. FPFH is known as very weak feature for matching scene level and local instances. You may visualize the actual FPFH matches to understand this better. I believe there should be too many false-positive matches that are fead to FGR. FPFH feature is not very admirable feature for this tough task, though FGR can be used with other feature.

10 of the scenes are from sceneNN and 2 from scanNet, the templates are taken from shapeNet.

I believe the registration task is not well-defined for FGR. The task you are trying is to register two different geometries that looks similar or came from same classes. If FGR is used with FPFH, they should come from the exactly the same geometry. As an alternative way, FGR can be used with semantic feature, but this is future work.

[YouYue123]

May I know what can be recommended for FPFH feature matching visualisation in order to debug this?