search

neka-nat / probreg

Python package for point cloud registration using probabilistic model (Coherent Point Drift, GMMReg, SVR, GMMTree, FilterReg, Bayesian CPD)
MIT License
855 stars 148 forks source link

Nonrigid cpd not functional #93

Closed a-parida12 closed 2 years ago

a-parida12 commented 2 years ago

I have been trying the cpd from the package it seems that the non-linear mode is not functional.

tf_param, _, _ = cpd.registration_cpd(source_pt, target_pt, maxiter=10000, tf_type_name='nonrigid')

Here is a screenshot of the output. Red is the source cloud, green the target cloud and blue the transformed cloud. Screenshot from 2022-07-11 09-31-59

I printed the iterations and res.q and q for your reference. Screenshot from 2022-07-11 09-24-26

In the screen shot below the tf_type_name='rigid' seems to be working fine. Screenshot from 2022-07-11 09-26-16

neka-nat commented 2 years ago

Thanks for trying probreg! NonrigidCPD has some parameters, so how about adjusting them? https://probreg.readthedocs.io/en/latest/probreg.html#probreg.cpd.ConstrainedNonRigidCPD