![]() |
ACloudViewer
3.9.4
A Modern Library for 3D Data Processing
|
#include <RobustKernel.h>


Public Member Functions | |
| HuberLoss (double k) | |
| Parametrized Constructor. More... | |
| double | Weight (double residual) const override |
Public Member Functions inherited from cloudViewer::pipelines::registration::RobustKernel | |
| virtual | ~RobustKernel ()=default |
Public Attributes | |
| double | k_ |
| Scaling paramter. More... | |
HuberLoss loss function used for outlier rejection.
The loss p(r) for a given residual 'r' is computed as follow: p(r) = r^2 for abs(r) <= k, p(r) = k^2 * (abs(r) - k/2) for abs(r) > k
For more information: http://en.wikipedia.org/wiki/Huber_Loss_Function
Definition at line 93 of file RobustKernel.h.
|
inlineexplicit |
Parametrized Constructor.
| k | Is the scaling paramter of the huber loss function. 'k' corresponds to 'delta' on this page: http://en.wikipedia.org/wiki/Huber_Loss_Function |
Definition at line 100 of file RobustKernel.h.
|
overridevirtual |
The weight w(r) for a given residual 'r' is computed as follow: w(r) = 1.0 for abs(r) <= k, w(r) = k / abs(r) for abs(r) > k Where k Is the scaling paramter of the loss function.
| residual | Residual value obtained during the optimization step. |
Implements cloudViewer::pipelines::registration::RobustKernel.
Definition at line 31 of file RobustKernel.cpp.
| double cloudViewer::pipelines::registration::HuberLoss::k_ |
Scaling paramter.
Definition at line 112 of file RobustKernel.h.
Referenced by cloudViewer::pipelines::registration::pybind_robust_kernels(), and Weight().