somanshu25 / gpu-accelerated-point-cloud-registration-using-hierarchical-gmm Goto Github PK

hi,
I had tried the python example: hgmm_cupy_new.py in my PC (WIN10 + PY36),
and add possible missing code in line357: n = len(momentsZero)
The log show like this:

geometry::PointCloud with 1406 points.
geometry::PointCloud with 1406 points.
Level: 0
The current IDx values are:
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
Likelihood value is: -48561.51961124496
The difference is the value is -48561.51961124496
Level: 0
The current IDx values are:
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
Likelihood value is: -48561.51961124496
The difference is the value is 0.0
break
Level: 1
The current IDx values are:
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
Likelihood value is: -48561.51961124496
The difference is the value is -48561.51961124496
Level: 1
The current IDx values are:
[8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8]
Likelihood value is: -48561.51961124496
The difference is the value is 0.0

However, the result shown in open3d, the red and green one seem not match
so I wonder that something wrong about this, because the likelihood value is weird
Could you check that for me? thanks!!

BR

I am glad to see your code, but when I tested it, I found that there is no implementation of the "pointCloudRegisterGPU" function. It is in /src/c++/gmm_registration/gmm_reg.cu. I think this should be the main part. Could you please provide it? Thank you!