Author Topic: Parameters meaning  (Read 3854 times)

Offline StanSmith

  • Newbie
  • *
  • Posts: 2
    • View Profile
Parameters meaning
« on: March 30, 2014, 02:17:10 PM »
Hey! Can you explain what parameters Graph density multiplier, Initial rigidness and Rigidness divisor do?
Don't mess with Stan Smith

Offline Andrew Krovopuskov

  • Administrator
  • Sr. Member
  • *****
  • Posts: 330
  • Technical support
    • View Profile
Re: Parameters meaning
« Reply #1 on: March 31, 2014, 02:36:53 PM »
Hey! Sure!

Let me explain a big picture.
1. Wrap has two models to work with: Template and Target.


2. First it subsamples Template model into a set of Control nodes.
Note that Mean vertex distance value. I will explain it later.


3. Wrap tries to move the nodes so that the Template model matches the Target.
Then it subsamples the Template again to create more nodes. Then it continue matching with new nodes. The process iterates until the maximum nodes dencity is achieved.
This maximum nodes density is specified as a fraction of Mean node dencity. And this value is exactly what parameter Graph density multiplier means!


If we set Graph density multiplier to 2.0 the result we get is:


If we set Graph density multiplier to 1.0 (nodes will have the same density as mean vertex density) we have result:


Why not to set Graph density multiplier to 1.0 all the time?
  • if the Template mesh is big it will take much time to compute
  • when you set Graph density multiplier > 1.0 the result will be more resistant to Target model noise. That is a desirable side-effect in many cases
In practice: always set it to 2.0. If you want better details on resulting topology, set it to 1.0 and lower.

Offline Andrew Krovopuskov

  • Administrator
  • Sr. Member
  • *****
  • Posts: 330
  • Technical support
    • View Profile
Re: Parameters meaning
« Reply #2 on: March 31, 2014, 02:51:58 PM »
As for Initial rigidness and Rigidness divisor:
Template model has a property of rigidness which reflects the ability of the model to bend. At the begining of optimization the rigidness is set to Initial rigidness value. With every iteration Wrap subdivides control nodes and reduces the rigidness of the model so it becomes more flexible
newRigidness = oldRigidness / rigidnessDivisor
The more the Rigidness divisor value the faster the model becomes flexible.

In pracice always set InitialRigidness to 3.0 and Rigidness divisor to 3.0.
If you feel that at the last iteration the Template model is not flexible enough to fit the Target properly, then set Rigidness divisor to 2.0-2.5
If Rigidness divisor is too big the model will become to flexible at the first iterations and that will lead to errors and irregular mesh.

Offline cateye999

  • Newbie
  • *
  • Posts: 1
    • View Profile
Re: Parameters meaning
« Reply #3 on: September 03, 2014, 12:10:40 PM »
Hey Andrew,

Since version has been upgraded and parameters also changed.
Any explanation for newer version? ex.  Final rigidness = Rigidness divisor? Good number for Max interations?
Thank you.

Offline Andrew Krovopuskov

  • Administrator
  • Sr. Member
  • *****
  • Posts: 330
  • Technical support
    • View Profile
Re: Parameters meaning
« Reply #4 on: September 04, 2014, 07:00:17 PM »
Hey!
Welcome to our forum! Let me prepare a detailed explanation for you but for now:

Initial density: 0.1 - 1.0
How dense are nodes at last iteration. 1.0 means that the nodes density will be equal to average vertex density of a model. The less density the slower calculations.

Density multiplier: 2.0 - 2.5
How fast nodes' density will grow with each iteration.

Initial rigidness: 2.0 - 3.0
How rigid is the base mesh at the beginning.

Final rigidness: 0.1 - 0.2
How rigid is the base mesh at the last iteration. The less final rigidness the better it fits the scan but small rigidness might cause irregular grid.

Max iterations: 20
Just a number of iterations in each step.

It's better to leave most parameters with default values. There're only two parameters that often require fine tuning:
Initial density (very slow and accurate / fast and inaccurate)
Final rigidness (fits scan details well but can produce irregular grid / fits only big details, insensitive to noise, mesh is more regular)