Object deformation network system and method

1. A computer-implemented method comprising:

receiving a set of input values for posing an initial mesh defining a surface of a three-dimensional object;

providing the set of input values to a neural network trained based upon training posed meshes generated using a rigging model to generate mesh offset values defined in three-dimensional space based upon the set of input values and the initial mesh, the neural network comprising an input layer, an output layer, and a plurality of intermediate layers;

generating, by the output layer of the neural network, a set of offset values corresponding to a set of three-dimensional target points based on the set of input values;

applying the offset values to the initial mesh to generate a posed mesh; and

outputting the posed mesh for generating an animation frame, wherein the neural network is trained by:

executing the rigging model based upon a set of training values to pose a training input mesh to generate a training deformed mesh;

providing the set of training values as input to the input layer of the neural network;

generating, based on output of the neural network, a training estimated mesh; and

updating parameters of the neural network to minimize a loss function between the training estimated mesh and the training deformed mesh.

2. The method of claim 1 , further comprising repeating the training using multiple training deformed meshes corresponding to two or more of:

mesh position values generated for an animation using the rigging model;

mesh position values generated by manipulating the training input mesh for a predetermined set of poses;

mesh position values generated by manipulating a rig-control variable between minimum and maximum values; and

mesh position values generated by manipulating a subset of the training input mesh using the rigging model.

3. The method of claim 1 , wherein the training input mesh is substantially the same as the initial mesh.

4. The method of claim 1 , wherein the intermediate layers of the neural network comprise one or more skip connections.

5. The method of claim 1 , wherein the points of the set of three-dimensional target points are generated substantially simultaneously.

6. The method of claim 1 , wherein:

an intermediate layer adjacent to the output layer generates a set of weights; and

the output layer applies the set of weights to a respective set of blendshapes for generating the offset values for the posed mesh, wherein the set of blendshapes are a subset of possible blendshapes selected using principal component analysis.

7. The method of claim 6 , wherein a width of the intermediate layer adjacent to the output layer is equal to a number of components used for the principal component analysis.

8. The method of claim 1 , wherein the method is performed in less than about 10 milliseconds.

9. The method of claim 1 , wherein the object is a character and wherein the initial mesh is a facial mesh defining a face of the character.

10. A system for performing animation, the system comprising one or more processors configured to:

receive a set of input values for posing an initial mesh defining a surface of a three-dimensional object;

provide the set of input values to a neural network trained based upon training posed meshes generated using a rigging model to generate mesh offset values defined in three-dimensional space based upon the set of input values and the initial mesh, the neural network comprising an input layer, an output layer, and a plurality of intermediate layers;

generate, by the output layer of the neural network, a set of offset values corresponding to a set of three-dimensional target points based on the set of input values;

applying the offset values to the initial mesh to generate a posed mesh; and

output the posed mesh for generating an animation frame, wherein:

an intermediate layer adjacent to the output layer generates a set of weights; and

the output layer applies the set of weights to a respective set of blendshapes for generating the offset values for the posed mesh.

11. The system of claim 10 , further comprising training the neural network by:

executing the rigging model based upon a set of training values to pose a training input mesh to generate a training deformed mesh;

providing the set of training values as input to the input layer of the neural network;

generating, based on output of the neural network, a training estimated mesh; and

updating parameters of the neural network to minimize a loss function between the training estimated mesh and the training deformed mesh.

12. The system of claim 11 , further comprising repeating the training using multiple training deformed meshes corresponding to two or more of:

mesh position values generated for an animation using the rigging model;

mesh position values generated by manipulating the training input mesh for a predetermined set of poses;

mesh position values generated by manipulating a rig-control variable between minimum and maximum values; and

mesh position values generated by manipulating a subset of the training input mesh using the rigging model.

13. The system of claim 11 , wherein the training input mesh is substantially the same as the initial mesh.

14. The system of claim 10 , wherein the intermediate layers of the neural network comprise one or more skip connections.

15. The system of claim 10 , wherein the points of the set of three-dimensional target points are generated substantially simultaneously.

16. The system of claim 10 , wherein:

the set of blendshapes are a subset of possible blendshapes selected using principal component analysis.

17. The system of claim 16 , wherein a width of the intermediate layer adjacent to the output layer is equal to a number of components used for the principal component analysis.

18. The system of claim 10 , wherein the receiving, providing, generating, and outputting are performed in less than about 10 milliseconds.

19. The system of claim 10 , wherein the object is a character and wherein the initial mesh is a facial mesh defining a face of the character.