Should it be lmd * np.linalg.inv(gmat) as in the paper？

neka-nat / probreg

Python package for point cloud registration using probabilistic model (Coherent Point Drift, GMMReg, SVR, GMMTree, FilterReg, Bayesian CPD)

MIT License

861 stars 148 forks source link

Closed sun6s closed 4 years ago

sun6s commented 4 years ago

https://github.com/neka-nat/probreg/blob/265658908b5857d497f3f37703acb6899a7bc63d/probreg/bcpd.py#L118 If we follow the paper,it shoud be sigma_mat_inv = lmd gmat + s2s2 np.diag(nu). However, if it is changed,the result is wrong. Is it wrong to use "np.linalg.inv(gmat)" in the paper?

neka-nat commented 4 years ago

Thank you! I've pushed fixes #48 and this issue to master. Fixing both is working fine for the bunny.