Candide3 face animation model

To learn about a model today, Candide3 face animation model.
​

Before understanding Candide, let's understand a standard of Mpeg4.

Mpeg is an image compression coding technology, Mpeg technology defines some standards.Some of them have developed standardized standards for 3D face animation.There are 2 main concepts to describe 3D face animation:

1. FDPS(Facial Definition Parameters)

   It is used to describe the face geometry and texture information.

2. FAPS(Facial Animation Parameters)

   It is used to describe the dynamic changes of the face.On the basis of the above-mentioned FDPS to generate personalized faces, the FAPS parameters are used to realize the expression animation deformation.For example, describe the rise or fall of the eyebrows, and different FAP combinations form a specific expression or facial animation

In short, FDP and FAP constitute a set of facial structure and expression animation, and the Candide 3 model is a subset of this set.

1. Candide 3 includes some FDP vertices and some FAP expression definitions, and discards minor expression changes such as ears and tongues.There are 113 feature points and 168 triangular patches, which constitute a 3D face model.
   ​
   

2. Candide 3 targets another face animation standard - FACS (Facial Action Coding System).

FACS defines a series of facial animation parameters AU based on the principles of facial muscle movement and expression control, and the correlation between AUs constitutes expressions.

The following table shows the AU parameters and their meanings in FACS:
​

The meaning of the AU parameter

The following table shows the combination of AU as an expression:
​

AU combination becomes expression
3. The Candide 3 model maps the image space by the following formula:
​

Candide model mapping formula

Where:

s is the final 3D face model with expressions;

a, P, t are parameters that control model scaling, rotation, and displacement;

S_0 is the standard face model;

S_1-n is the face animation parameter;

w_1-n is S_1-n is animation parameter weight;

n Number of animation parameters.

The animation parameters here are from the facial expression animation unit AU.

Given a photo, we can determine the 3D candide face model according to the feature points on the 2D face photo. At the same time, we can control the facial expression animation through the AU parameters, that is, control the spatial displacement of the 3D model vertices.

Prepare a photo of Julie below:
​

Generate a personalized face, add a grinning animation to the model (click to enlarge if you can't see clearly):
​

Draw mouth animation:
​
reverse directionPout
​
Summary: Candide3 is a face animationModel.Its data contains a standard face model - vertex and texture information, and a series of animation units - a combination of vertex displacements based on muscle control and expression principles.
The middle face model is customized through face feature points to generate a personalized face model. At the same time, the animation parameters provided by AU, together with other parameters such as rotation and scaling, can be used to change the spatial position of model vertices to form expressions and animations.
​
Face animation generation process