Principal Component Analysis Node (PCA)...

   Views 795   Replies 2   Subscribers 0
User Avatar
Member
79 posts
Joined: Aug. 2013
Offline
Does anyone know how to use this node? What type of input does it expect?
User Avatar
Member
172 posts
Joined: May 2021
Offline
coccarolla
Does anyone know how to use this node? What type of input does it expect?

PCA is a common technique used in data-driven Character Customization Tools etc. Here is a description standard model from the ICT-Lab:

What is a Linear 3D Morphable Face Model?
https://www.youtube.com/watch?v=MlGkzFeyCYc [www.youtube.com]

This node has three modes:

Analysis
- takes in samples of geometry (data) with the same # points but deformed over time or variants of shape
- computes Principal components for the number of components you specify

Project - Projects sample of geometry into the PCA constructed from the Analysis, returning coefficients
Reconstruct - Constructs geometry from coefficients the PCA analysis defined, returning geometry

It is like finding the parameters to slide to change a geometry based on understanding the right parts of the geometry to influence.

It is demoed in the MLDeformer project on the SideFX example project site here:

https://www.sidefx.com/contentlibrary/ml-deformer/ [www.sidefx.com]
Edited by PHENOMDESIGN - July 23, 2024 13:02:02
PHENOM(enological) DESIGN;
Experimental phenomenology (study of experience) is a category of philosophy evidencing intentional variations of subjective human experiencing where both the independent and dependent variable are phenomenological. Lundh 2020
User Avatar
Member
79 posts
Joined: Aug. 2013
Offline
PHENOMDESIGN
coccarolla
Does anyone know how to use this node? What type of input does it expect?

PCA is a common technique used in data-driven Character Customization Tools etc. Here is a description standard model from the ICT-Lab:

What is a Linear 3D Morphable Face Model?
https://www.youtube.com/watch?v=MlGkzFeyCYc [www.youtube.com]

This node has three modes:

Analysis
- takes in samples of geometry (data) with the same # points but deformed over time or variants of shape
- computes Principal components for the number of components you specify

Project - Projects sample of geometry into the PCA constructed from the Analysis, returning coefficients
Reconstruct - Constructs geometry from coefficients the PCA analysis defined, returning geometry

It is like finding the parameters to slide to change a geometry based on understanding the right parts of the geometry to influence.

It is demoed in the MLDeformer project on the SideFX example project site here:

https://www.sidefx.com/contentlibrary/ml-deformer/ [www.sidefx.com]

Hey, thanks for the detailed answer. This node does an enormity of an entirely different thing than what I thought it did.
  • Quick Links