On-set facial performance capture and transfer to a three-dimensional computer-generated model

1. A method of transferring a facial expression from a subject to a computer generated character, the method comprising:

receiving a plate with an image of the facial expression of the subject where the plate comprises more than a million pixels with each pixel having a particular RGB value, a three-dimensional parameterized deformable model of a face of the subject where different facial expressions of the subject can be obtained by varying values of model parameters, a model of a camera rig used to capture the plate, and a virtual lighting model that estimates lighting conditions when the image of the plate was captured;

solving for the facial expression in the plate by executing a deformation solver to solve for at least some parameters of the three-dimensional parameterized deformable model with a differentiable renderer and shape from shading techniques, using, as inputs, the three-dimensional parameterized deformable model, the model of the camera rig and the virtual lighting model over a series of iterations to infer geometry of the facial expression in the plate and generate a final facial mesh using a set of parameter values of the three-dimensional parameterized deformable model which result in a final facial expression that closely matches the facial expression of the subject in the plate; and

wherein, during each iteration of the series of iterations, the differentiable renderer generates a rendering of the three-dimensional parameterized deformable model and the deformation solver minimizes differences between RGB values of the plate and RGB values of corresponding pixels in the rendered version of the three-dimensional parameterized deformable model.

2. The method of claim 1 , wherein, the three-dimensional parameterized deformable model comprises a plurality of blendshapes representing different facial expressions of the subject and wherein, the set of parameter values of the three-dimensional parameterized deformable model for the final facial mesh comprises a set of blendshape weight values, one blendshape weight value per each blendshape.

3. The method of claim 2 , wherein the set of parameter values of the three-dimensional parameterized deformable model for the final facial mesh further comprises rotation and translation values that represent a rigid adjustment of a head of the subject.

4. The method of claim 3 , wherein, the computer-generated character includes a head that is sized or shaped differently than the head of the subject and the set of parameter values of the three-dimensional parameterized deformable model for the final facial mesh further comprises a value for a delta vector that represents a per vertex displacement used in transferring the facial expression of the subject to the computer-generated character.

5. The method of claim 2 , further comprising, generating a visual effect from a set of weighted blendshapes and incorporating the visual effect into one or more of an animation, a movie, a video, a video game, or virtual or augmented reality content.

6. The method of claim 1 , wherein, an initial iteration of the solving for the facial expression in the plate comprises:

rendering an initial facial mesh, generated from the three-dimensional parameterized deformable model, representing a neutral expression of the subject;

trying to minimize differences between the RGB values of the plate and RGB values of the rendered initial facial mesh representing the neutral expression; and

generating an updated facial mesh in which the facial expression of the subject more closely matches the facial expression of the subject in the plate than does the neutral expression of the initial facial mesh.

7. The method of claim 6 , wherein, for each iteration of the solving step, an output of current iteration is closer to an actual representation of the subject in the plate than an output of the previous iteration.

8. The method of claim 6 , wherein, solving for the facial expression in the plate includes executing a plurality of different deformation solvers, wherein each deformation solver executes a plurality of iterations before next solver is executed and wherein each deformation solver has at least one cost function associated with it that defines an objective that a respective deformation solver tries to minimize when executing the respective deformation solver.

9. The method of claim 8 , wherein, the plurality of different deformation solvers are executed in a predetermined sequence that is defined by a recipe selected from a plurality of predetermined recipes stored in a library.

10. The method of claim 9 , wherein, each predetermined recipe in the plurality of predetermined recipes includes one or more deformation solvers, each of which has at least one cost function associated with it.

11. The method of claim 1 , wherein, the model of the camera rig is match-moved and models a taking camera and at least one witness camera mounted on the same camera rig.

12. The method of claim 11 , wherein, the image of the facial expression of the subject includes a small set of markers placed on rigid portions of the face of the subject.

13. The method of claim 12 , wherein, data representing positions of the set of markers are used during the solving step for determining rotation and translation of a skull of the subject.

14. The method of claim 1 , wherein, the plate is captured from cameras mounted on a rig within a performance area separate from the subject.

15. The method of claim 1 , wherein, the three-dimensional parameterized deformable model includes rotational and translations joints in addition to blendshapes.

16. A method of transferring a facial expression from a subject during a performance to a computer generated character, the method comprising:

obtaining digital video footage of the performance including a plurality of sequentially ordered plates, each of which includes an image of the facial expression of the subject during the performance and comprises more than a million pixels with each pixel having a particular RGB value;

obtaining a three-dimensional parameterized deformable model of a face of the subject where different facial expressions of the subject can be obtained by varying the values of model parameters, a model of a camera rig used to capture the performance, and a virtual lighting model that estimates lighting conditions used during the performance;

generating a computer model of the performance by, for each individual plate in the plurality of sequentially ordered plates, processing a respective individual plate independently of other plates in the plurality of sequentially ordered plates to solve for the facial expression in the respective plate with a differential renderer and shape from shading techniques, using, as inputs, the three-dimensional parameterized deformable model, the model of the camera rig and the virtual lighting model over a series of iterations to infer geometry of the facial expression in the respective plate and generate a final facial mesh using a set of parameter values for the three-dimensional parameterized deformable model which result in a final facial expression that closely matches the facial expression of the subject in the respective plate being processed; and

wherein, during each iteration of the series of iterations, the differentiable renderer generates a rendering of the three-dimensional parameterized deformable model and the deformation solver minimizes differences between RGB values of the respective plate and RGB values of corresponding pixels in the rendered version of the three-dimensional parameterized deformable model.

17. The method of claim 16 , wherein, the three-dimensional parameterized deformable model comprises a plurality of blendshapes representing different facial expressions of the subject and wherein, the set of parameter values comprises a set of blendshape weight values, one blendshape weight value per each blendshape.

18. The method of claim 17 , wherein, the set of parameter values further comprises rotation and translation values that represent a rigid adjustment of a head of the subject.

19. The method of claim 18 , wherein

an initial iteration of the solving for the facial expression in the respective plate comprises: (i) rendering an initial facial mesh, generated from the three-dimensional parameterized deformable model, representing a neutral expression of the subject; (ii) trying to minimize differences between the RGB values of the respective plate and RGB values of the rendered initial facial mesh representing the neutral expression; and (iii) generating an updated facial mesh in which the facial expression of the subject more closely matches the facial expression of the subject in the respective plate than does the neutral expression of the initial facial mesh; and

wherein, for each additional iteration of the solving step, an output of current iteration is closer to an actual representation of the subject in the respective plate than an output of the previous iteration.