Compressed representations for appearance of fiber-based digital assets

Patent metadata

Field	Value
Publication number	US-12561897-B2
Application number	US-202318354855-A
Country	US
Kind code	B2
Filing date	Jul 19, 2023
Priority date	Jul 19, 2023
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

Techniques for generation of compressed representations for appearance of fiber-based digital assets are described that support computationally efficient and high fidelity rendering of digital assets that include fiber primitives under a variety of lighting conditions and view directions. A processing device, for instance, receives a digital asset that includes fiber primitives to be included in a three-dimensional digital scene. The processing device generates a compressed representation of the digital asset that maintains a geometry of the digital asset and includes a precomputed light transport. The processing device then inserts the compressed representation into the digital scene, such as at a location relative to one or more light sources. The processing device applies one or more lighting effects to the compressed representation based on the precomputed light transport and the location relative to the one or more light sources.

First claim

What is claimed is:

1 . A method comprising:
receiving, by a processing device, a digital asset to be included in a three-dimensional digital scene, the digital asset including fiber primitives; generating, by the processing device, a training dataset that includes training images depicting the digital asset, the training dataset including one or more training samples per pixel that depicts the digital asset within the training images, a respective training sample of a respective pixel including one or more parameters that are specific to an individual fiber primitive on which the respective pixel is depicted; generating, by the processing device, a compressed representation of the digital asset that includes a precomputed light transport for the fiber primitives at least partially encoded into a neural model that includes a multilayer perceptron and a feature grid corresponding to a cube of data spatially aligned with cylindrical representations of the fiber primitives, the compressed representation generated by jointly training the feature grid and the multilayer perceptron using the training dataset; inserting, by the processing device, the compressed representation of the digital asset into a location relative to a digital light source within the three-dimensional digital scene; and applying, by the processing device, one or more lighting effects to the fiber primitives based on the precomputed light transport and the location relative to the digital light source.

2 . The method as described in claim 1 , wherein the applying the one or more lighting effects is based on a view direction of the compressed representation within the three-dimensional digital scene.

3 . The method as described in claim 1 , wherein the applying the one or more lighting effects includes determining an intersection point of a ray traced from a virtual camera that defines a view direction with a fiber primitive of the compressed representation and computing a color value to apply to the fiber primitive at the intersection point based on the precomputed light transport.

4 . The method as described in claim 3 , wherein the compressed representation includes the feature grid that parameterizes a geometry of the cylindrical representations of the fiber primitives and the multilayer perceptron, and wherein the applying the one or more lighting effects includes extracting a feature vector from the feature grid based on the intersection point and evaluating the feature vector using the multilayer perceptron to determine the color value.

5 . The method as described in claim 4 , wherein the feature vector is combined with the view direction, a light direction, a fiber direction, and a fiber axis offset for the intersection point for input to the multilayer perceptron.

6 . The method as described in claim 1 , wherein the training images depict the digital asset viewed from a plurality of different locations within the three-dimensional digital scene.

7 . The method as described in claim 6 , wherein generating the compressed representation includes utilizing the training dataset to adjust weights of the multilayer perceptron of the compressed representation and update the feature grid of the compressed representation.

8 . The method as described in claim 1 , wherein the precomputed light transport represents a scattering interaction of light rays that interact with the fiber primitives.

9 . The method as described in claim 1 , wherein the applying the one or more lighting effects includes applying RGB color values to points along the fiber primitives.

10 . The method as described in claim 1 , wherein the feature grid is defined by a UVT coordinate system, a U-axis and a V-axis of the UVT coordinate system representing a flattened structure of the digital asset and a T-axis of the UVT coordinate system representing a length of the cylindrical representations of the fiber primitives.

11 . The method as described in claim 1 , wherein the one or more parameters include that are specific to the individual fiber primitive include one or more of:
an intersection point at which a ray traced from a virtual camera intersects with the individual fiber primitive; a fiber axis offset defining a first distance from a center of the individual fiber primitive to the intersection point; and a second distance from a base of the individual fiber primitive to the intersection point.

12 . The method as described in claim 1 , wherein jointly training the feature grid and the multilayer perceptron using the training dataset includes:
receiving a ground truth radiance for the respective pixel; receiving the respective training sample of the respective pixel; generating a predicted radiance for the respective pixel based on the one or more parameters that are specific to the individual fiber primitive on which the respective pixel is depicted; and adjusting parameters of the feature grid and the multilayer perceptron based on the ground truth radiance and the predicted radiance.

13 . A system comprising:
a memory component; and a processing device coupled to the memory component, the processing device to perform operations comprising:
receiving a digital asset to be included in a three-dimensional digital scene, the digital asset including fiber primitives;
generating a training dataset that includes training images depicting the digital asset, the training dataset including one or more training samples per pixel that depicts the digital asset within the training images, a respective training sample of a respective pixel including one or more parameters that are specific to an individual fiber primitive on which the respective pixel is depicted;
generating a compressed representation of the digital asset that includes a precomputed light transport for the fiber primitives, the precomputed light transport at least partially represented by a neural model that includes a multilayer perceptron and a feature grid corresponding to a cube of data aligned with cylindrical representations of the fiber primitives, the compressed representation generated by jointly training the feature grid and the multilayer perceptron using the training dataset;
inserting the compressed representation of the digital asset into a location relative to a digital light source within the three-dimensional digital scene; and
applying one or more lighting effects to the fiber primitives based on the precomputed light transport and the location relative to the digital light source.

14 . The system as described in claim 13 , wherein the applying the one or more lighting effects includes determining an intersection point of a ray traced from a virtual camera that defines a view direction with a fiber primitive of the compressed representation and computing a color value to apply to the fiber primitive at the intersection point based on the precomputed light transport.

15 . The system as described in claim 14 , wherein the compressed representation includes the feature grid that parameterizes a geometry of the cylindrical representations of the fiber primitives and the multilayer perceptron, and wherein the applying the one or more lighting effects includes extracting a feature vector from the feature grid based on the intersection point and evaluating the feature vector using the multilayer perceptron to determine the color value.

16 . The system as described in claim 15 , wherein the feature vector represents the view direction, a light direction, a fiber direction, and a fiber axis offset for the intersection point for input to the multilayer

Assignees

• Adobe Inc

Inventors

• Lakshminarayana Krishna Bhargava Mullia
• Sun Xin
• Hašan Miloš
• Luan fujun

Classifications

• G06T15/06

  Ray-tracing · CPC title

• G06T2210/56

  Particle system, point based geometry or rendering · CPC title

• G06T2210/21

  Collision detection, intersection · CPC title

• G06T15/005

  General purpose rendering architectures · CPC title

• G06T15/506 Primary

  Illumination models · CPC title