Preserving geometry details in a sequence of tracked meshes

1. An electronic device, comprising:

circuitry configured to:

acquire a set of three-dimensional (3D) scans of an object of interest;

acquire a sequence of tracked meshes of the object of interest,

wherein the acquired sequence of tracked meshes comprises a set of tracked meshes which temporally corresponds to the acquired set of 3D scans;

generate a set of displacement maps based on differences between the set of tracked meshes and the acquired set of 3D scans;

compute a plurality of vectors, each of which includes surface tension values associated with mesh vertices of a corresponding tracked mesh of the sequence of tracked meshes;

train a neural network model on a task of displacement map generation, based on the computed set of displacement maps and a corresponding set of vectors of the computed plurality of vectors;

apply the trained neural network model on the computed plurality of vectors to generate a plurality of displacement maps; and

update each tracked mesh of the acquired sequence of tracked meshes based on a corresponding displacement map of the generated plurality of displacement maps.

2. The electronic device according to claim 1 , wherein the circuitry is further configured to:

synchronize a second imaging device with a first imaging device;

after the synchronization, control the first imaging device to capture a first sequence of image frames of the object of interest in a dynamic state; and

control the second imaging device to capture a second sequence of image frames of the object of interest in the dynamic state.

3. The electronic device according to claim 2 , wherein the first imaging device is a digital still camera and the second imaging device is a video camera, and

the first imaging device captures the first sequence of image frames at a frame rate that is less than a frame rate at which the second imaging device captures the second sequence of image frames.

4. The electronic device according to claim 2 , wherein an image resolution of each of the first sequence of image frames is greater than an image resolution of each image frame of the second sequence of image frames.

5. The electronic device according to claim 2 , wherein the circuitry is further configured to execute a first set of operations including a photogrammetry operation to acquire the set of 3D scans, and

wherein the execution is based on the captured first sequence of image frames.

6. The electronic device according to claim 2 , wherein the circuitry is further configured to execute a second set of operations including a mesh-tracking operation to acquire the sequence of tracked meshes, and

wherein the execution is based on the captured second sequence of image frames and a parametric 3D model of the object of interest.

7. The electronic device according to claim 1 , wherein a polycount in each tracked mesh of the acquired sequence of tracked meshes is less than a polycount in each 3D scan of the acquired set of 3D scans.

8. The electronic device according to claim 1 , wherein the circuitry is further configured to construct a training dataset to include pairs of input-output values, each of which comprises of:

a vector of the computed set of vectors corresponding to the set of tracked meshes as an input for the neural network model, and

a displacement map of the generated set of displacement maps as a ground truth for an output of the neural network model,

wherein the neural network model is trained for a number for epochs on the pairs of input-output values until a loss between the ground truth and the output of the neural network model is below a threshold.

9. The electronic device according to claim 1 , wherein the object of interest is a face portion of a person.

10. The electronic device according to claim 9 , wherein the circuitry is further configured to:

determine one or more first tracked meshes of the sequence of tracked meshes to be associated with a neutral face expression;

determine the surface tension values associated with the mesh vertices of each of the determined one or more first tracked meshes as zero;

determine one or more second tracked meshes of the sequence of tracked meshes to be associated with a face expression different from the neutral face expression;

compare the mesh vertices of each of the determined one or more second tracked meshes with reference values of the mesh vertices for the neutral face expression; and

determine the surface tension values associated with the mesh vertices of each of the determined one or more second tracked meshes, based on the comparison.

11. The electronic device according to claim 1 , wherein the circuitry is further configured to resample each displacement map of the generated plurality of displacement maps until a resolution of each resampled displacement map of the generated plurality of displacement maps corresponds to a polycount of a 3D scan of the acquired set of 3D scans.

12. The electronic device according to claim 11 , wherein the update comprises of:

a first application of a resampling operation on each tracked mesh of the sequence of tracked meshes until a polycount of each resampled mesh of the resampled sequence of tracked meshes matches that of a corresponding 3D scan of the set of 3D scans; and

a second application of each resampled displacement map of the resampled set of displacement maps to a corresponding resampled mesh of the resampled sequence of tracked meshes, to obtain the updated sequence of tracked meshes.

13. The electronic device according to claim 1 , wherein the updated sequence of tracked meshes correspond to a set of blend-shapes for animation.

14. A method, comprising:

acquiring a set of three-dimensional (3D) scans of an object of interest;

acquiring a sequence of tracked meshes of the object of interest,

wherein the acquired sequence of tracked meshes comprises a set of tracked meshes which temporally corresponds to the acquired set of 3D scans;

generating a set of displacement maps based on differences between the set of tracked meshes and the acquired set of 3D scans;

computing a plurality of vectors, each of which includes surface tension values associated with mesh vertices of a corresponding tracked mesh of the sequence of tracked meshes;

training a neural network model on a task of displacement map generation, based on the computed set of displacement maps and a corresponding set of vectors of the computed plurality of vectors;

applying the trained neural network model on the computed plurality of vectors to generate a plurality of displacement maps; and

updating each tracked mesh of the acquired sequence of tracked meshes based on a corresponding displacement map of the generated plurality of displacement maps.

15. The method according to claim 14 , further comprising:

synchronizing a second imaging device with a first imaging device;

after the synchronization, controlling the first imaging device to capture a first sequence of image frames of the object of interest in a dynamic state; and

controlling the second imaging device to capture a second sequence of image frames of the object of interest in the dynamic state.

16. The method according to claim 15 , further comprising executing a first set of operations including a photogrammetry operation to acquire the set of 3D scans, wherein the execution is based on the captured first sequence of image frames.

17. The method according to claim 15 , further comprising executing a second set of operations including a mesh-tracking operation to acquire the sequence of tracked meshes, wherein the execution is based on the captured second sequence of image frames and a parametric 3D model of the object of interest.

18. The method according to claim 14 , further comprising constructing a training dataset to include pairs of input-output values, each of which comprises of:

a vector of the computed set of vectors corresponding to the set of tracked meshes as an input for the neural network model, and

a displacement map of the generated set of displacement maps as a ground truth for an output of the neural network model,

wherein the neural network model is trained for a number for epochs on the pairs of input-output values until a loss between the ground truth and the output of the neural network model is below a threshold.

19. The method according to claim 14 , further comprising:

determining one or more first tracked meshes of the sequence of tracked meshes to be associated with a neutral face expression;

determining the surface tension values associated with the mesh vertices of each of the determined one or more first tracked meshes as zero;

determining one or more second tracked meshes of the sequence of tracked meshes to be associated with a face expression different from the neutral face expression;

comparing the mesh vertices of each of the determined one or more second tracked meshes with reference values of the mesh vertices for the neutral face expression; and

determining the surface tension values associated with the mesh vertices of each of the determined one or more second tracked meshes, based on the comparison.

20. A non-transitory computer-readable medium having stored thereon computer implemented instructions that, when executed by an electronic device, causes the electronic device to execute operations, the operations comprising:

acquiring a set of three-dimensional (3D) scans of an object of interest;

acquiring a sequence of tracked meshes of the object of interest,

wherein the acquired sequence of tracked meshes comprises a set of tracked meshes which temporally corresponds to the acquired set of 3D scans;

generating a set of displacement maps based on differences between the set of tracked meshes and the acquired set of 3D scans;

computing a plurality of vectors, each of which includes surface tension values associated with mesh vertices of a corresponding tracked mesh of the sequence of tracked meshes;

training a neural network model on a task of displacement map generation, based on the computed set of displacement maps and a corresponding set of vectors of the computed plurality of vectors;

applying the trained neural network model on the computed plurality of vectors to generate a plurality of displacement maps; and

updating each tracked mesh of the acquired sequence of tracked meshes based on a corresponding displacement map of the generated plurality of displacement maps.