DETAILED ACTION
Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
2. The information disclosure statement (IDS) submitted on February 21, 2026 is considered by the examiner.
Response to Amendment
4. The amendment filed March 12, 2026 has been entered. Claims 1-10, 12-16, 18, 20, and 22-24 remain pending in the application.
Response to Arguments
5. Applicant's arguments filed March 12, 2026 have been fully considered but they are not persuasive.
6. Applicant argues that Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, Rainer et al. (“Neural BTF Compression and Interpolation”), hereinafter referred to as Rainer, and Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, do not teach "precomputed light transport computed for the digital asset independently of the digital scene to represent how light interacts with and/or propagates through the three-dimensional geometry".
Examiner replies that the Applicant’s arguments with respect to claim(s) 1 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Instead, Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, is used to teach the amended limitation as rejected below.
7. Applicant argues that it would not be obvious to combine Gupta and Rainer because Gupta teaches away. The Applicant asserts that Gupta designed its triplane and color prediction MLP to exclude lighting info.
Examiner replies that the Applicant’s arguments with respect to claim(s) 1 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Claim 1 no longer combines Rainer with Gupta but instead combines Yu with Gupta.
However, in response to the Applicant’s arguments against Gupta specifically, the Examiner replies that claim 1 does not require that the lighting be included in the triplane representation. Claim 1 only requires that the triplane representation is of the three-dimensional geometry.
Furthermore, in response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e., "encoding lighting into the triplane and MLP architecture") are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993).
Under the Examiner’s best understanding, Claim 1 only asserts that the neural model has the triplane and machine learning model and does not specify what exactly the triplane and machine learning model do as part of the neural model. Thus, claim 1 does not claim that the triplane and MLP architecture encode lighting information.
In addition, Gupta is not used to teach rendering the lighting effects using the MLP. Instead, Yu teaches rendering the object with the lighting effects using the encoded precomputed light transport and MLP. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
8. Applicant argues that the prior art does not teach the functionality and therefore does not realize the advantages of “generating, by the processing device, a compressed representation of the digital asset that encodes a precomputed light transport into a neural model having a triplane representation of the three-dimensional geometry and a machine learning model, the precomputed light transport computed for the digital asset independently of the digital scene to represent how light interacts with and/or propagates through the three-dimensional geometry” such as “supporting high fidelity shading operations for three-dimensional objects ‘while conserving computational resources relative to conventional techniques’”.
Examiner replies that with the new reference, Sheppard in view of Yu and Gupta do teach “generating, by the processing device, a compressed representation of the digital asset that encodes a precomputed light transport into a neural model having a triplane representation of the three-dimensional geometry and a machine learning model, the precomputed light transport computed for the digital asset independently of the digital scene to represent how light interacts with and/or propagates through the three-dimensional geometry” as rejected below.
Furthermore, in response to applicant's argument that the prior art cannot realize the advantages of “supporting high fidelity shading operations for three-dimensional objects ‘while conserving computational resources relative to conventional techniques’”, the fact that the inventor has recognized another advantage which would flow naturally from following the suggestion of the prior art cannot be the basis for patentability when the differences would otherwise be obvious. See Ex parte Obiaya, 227 USPQ 58, 60 (Bd. Pat. App. & Inter. 1985).
MPEP 2144(IV) also states that "the reason or motivation to modify the reference may often suggest what the inventor has done, but for a different purpose or to solve a different problem. It is not necessary that the prior art suggest the combination to achieve the same advantage or result discovered by applicant." Thus, the Applicant's arguments that the prior art does not teach the advantage the Inventor discovered are unpersuasive.
Conclusion: Thus, claims 1-10, 12-16, 18, 20, and 22-24 stand rejected.
Claim Objections
9. Claim 1 objected to because of the following informalities: Line 9 "and/or" should be just "and" or "or". Appropriate correction is required.
Claim 10 objected to because of the following informalities: Line Appropriate correction is required.
Claim 18 objected to because of the following informalities: Line Appropriate correction is required.
Claim Rejections - 35 USC § 112
10. The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
11. Claims 1-10, 12-16, 18, 20, and 22-24 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claims 1, lines 5-10 claim “generating, by the processing device, a compressed representation of the digital asset that encodes a precomputed light transport into a neural model having a triplane representation of the three-dimensional geometry and a machine learning model”.
The limitation is written in a confusing manner where the relation is unclear between the compressed representation, precomputed light transport, machine learning model, triplane representation, and machine learning model. It is unclear whether the neural model creates the compressed representation or if it is part of the process of creating a precomputed light transport. It also unclear whether the neural model is the combination of the triplane representation and machine learning model or if the precomputed light transport is the combination of the triplane representation and machine learning model. Thus, the scope and subject matter of claim 1 and its dependents are unclear.
The examiner will interpret the claim as such: generating, by the processing device, a compressed representation of the digital asset that encodes a precomputed light transport using a neural model; wherein the neural model includes a triplane representation of the three-dimensional geometry and a machine learning model; and wherein the precomputed light transport is computed for the digital asset independently of the digital scene to represent how light interacts with or propagates through the three-dimensional geometry.
The Applicant is also advised that amending the claim as above will overcome the noted rejection.
Regarding claim 10 and 18, both claims are also rejected for the same reasons as claim 1 above.
Claims 2-9, 12-16, 20, and 22-24 are also rejected by dependency on claims 1, 10, and 18.
Claim Rejections - 35 USC § 103
12. In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
13. The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
14. Claim(s) 1-2, 4-5, 9-10, 18, and 22-23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, and Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta.
15. Regarding claim 1, Sheppard teaches a method comprising: receiving, by a processing device, a digital asset defined by a three-dimensional geometry to be included in a digital scene, the digital scene including at least one digital scene element (Paragraph 87 and Figure 2 teaches the online processing unit 103 receiving a digital asset from the object data library 450; Paragraph 93 teaches including the digital asset from the object data library into a virtual environment or digital scene. The digital scene can be populated with multiple objects from the object library which satisfies including at least one digital scene element);
generating, by the processing device, a compressed representation of the digital asset (Paragraph 11 teaches providing objects or digital assets in a compressed format)
deploying, by the processing device, the compressed representation into the digital scene at a location relative to the at least one digital scene element (Paragraph 11 teaches the virtual environment comprises of a plurality of objects; Paragraph 97 teaches displaying an object from the object library at a particular position. The position is a location relative to other objects because there are other objects in the scene);
and rendering, by the processing device, the digital asset by applying one or more lighting effects to the three-dimensional geometry (Paragraph 174 teaches rendering the 3D object with light effects like shading and highlighting) (Paragraph 11 teaches the virtual environment comprises of a plurality of objects; Paragraph 97 teaches displaying an object from the object library at a particular position. The position is a location relative to other objects because there are other objects in the scene).
However, Sheppard is not relied upon for the below claim language: generating the compressed representation of the digital asset that encodes a precomputed light transport into a neural model having a triplane representation of the three-dimensional geometry and a machine learning model, the precomputed light computed for the digital asset independently of the digital scene to represent how light interacts with and/or propagates through the three-dimensional geometry; and rendering, by the processing device, the digital asset by applying one or more lighting effects to the three-dimensional geometry using the neural model based on the precomputed light transport.
Yu teaches generating the compressed representation of the digital asset that encodes a precomputed light transport into a neural model (Section 1 teaches an Object-Centric Neural Scattering Functions (OSFs) that approximates “the cumulative radiance transfer function, which models radiance transfer from a distant light to any outgoing direction at any spatial location for an object”. This teaches encoding the precompute light transport for a digital asset; Section 3.1 teaches an OSF which creates a compressed representation of the digital asset encoding the precomputed light transport into a neural model, seen in Equation 2; Page 18, Supplementary Material Section B teaches the neural model uses a multilayer perceptron. The multilayer perceptron teaches the neural model has a machine learning model),
the precomputed light transport computed for the digital asset independently of the digital scene to represent how light interacts with and/or propagates through the three-dimensional geometry (Section 3.1 and Figure 2 (third image) teach the precomputed light transport is computed independently of the digital scene using w_light and w_out. This represents how light interacts and propagates through the three-dimensional geometry of the object or digital asset. It also teaches the OSFs “model Lout as the scattered radiance due to external light sources”);
and rendering, by the processing device, the digital asset by applying one or more lighting effects to the three-dimensional geometry using the neural model based on the precomputed light transport and the location relative to the at least one digital scene element (Figure 1 teaches a digital asset added and rendered in a scene in the ‘Scene Composition’ image. The ‘Relighting’ image also teaches the digital asset can be rendered with different light effects applied to the three-dimensional geometry using the neural model taught by OSF. Section 3.1 teach the OSF has a neural model based on the precomputed light transport; Section 3.6 and Figure 3 teach using the neural model or learned OSF for scene composition. Figure 3 also teaches the neural model uses the location of the object relative to at least another digital scene element in order to sample shadow rays).
Sheppard and Yu are considered analogous to the claimed invention as because both are in the same field of rendering 3D objects in a scene. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of rendering the digital asset in a scene taught by Sheppard with the compressed representation of the precomputed light transport taught by Yu in order to allow for free-viewpoint relight and scene composition of both opaque and translucent objects (Yu Section 1 Paragraph 8).
However, Sheppard and Yu are not relied upon for the below claim language: the neural model having a triplane representation of the three-dimensional geometry.
Gupta teaches the neural model having a triplane representation of the three-dimensional geometry (Section 2.1 Paragraph 1 teaches Figure 1 uses neural fields that combines trilinear feature interpolation with a MLP decoder. Thus, the neural model has a triplane representation; Section 2.2 and Figure 1 teach an autoencoder that has a triplane representation of the three-dimensional geometry).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu with the triplane representation and machine learning models taught by Gupta in order to allow for quick generation of textured 3D objects (Gupta Abstract).
16. Regarding claim 2, Sheppard in view of Yu and Gupta teaches the limitations of claim 1. However, Sheppard and Yu are not relied upon for the below claim language: the method wherein the triplane representation includes feature grids that correspond to dimensions of the three- dimensional geometry to represent the digital asset.
Gupta teaches the method wherein the triplane representation includes feature grids that correspond to dimensions of the three-dimensional geometry to represent the digital asset (Section 2.2 and Figure 1 teach the process which converts a 3D object into a triplane representation. Figure 1 shows the feature grids that corresponds to the three-dimensional geometry. This results in the h_xz, h_xy, and h_yz triplane latent features which represent the XZ, XY, and YZ planes or feature grids).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu with the triplane representation taught by Gupta in order to allow for quick generation of textured 3D objects (Gupta Abstract).
17. Regarding claim 4, Sheppard in view of Yu and Gupta teaches the limitations of claim 1. However, Sheppard, and Gupta are not relied upon for the below claim language: the method wherein rendering the digital asset includes determining an intersection point of a ray traced from a virtual camera that defines a view direction with the compressed representation and computing a color value to apply to the compressed representation at the intersection point based on the precomputed light transport.
Yu teaches the method wherein rendering the digital asset includes determining an intersection point of a ray traced from a virtual camera that defines a view direction with the compressed representation and computing a color value to apply to the compressed representation at the intersection point based on the precomputed light transport (Section 3.3 teaches modeling opaque objects and determining an intersection point of a ray through x_surf. It teaches “x_surf denotes the first surface point hit by the camera ray r”. The camera ray ‘r’ teaches the ray traced from the virtual camera that defines a view direction; Section 3.4 teaches the color is then computed by mapping “camera ray radiance to pixel color”. Thus, the radiance calculated in Equation 4 teaches computing a color value to apply to the compressed representation at the intersection point based on the precomputed light transport).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of rendering the digital asset in a scene taught by Sheppard in view of Gupta with determining an intersection point to compute a color value taught by Yu in order to allow for free-viewpoint relight and scene composition of both opaque and translucent objects (Yu Section 1 Paragraph 8).
18. Regarding claim 5, Sheppard in view of Yu and Gupta teaches the limitations of claim 4. However, Sheppard and Yu are not relied upon for the below claim language: the method wherein the machine learning model is a multilayer perceptron, and rendering the digital asset includes: generating an input vector that corresponds to the intersection point based on the triplane representation; and evaluating the input vector using the multilayer perceptron to determine the color value.
Gupta teaches the method wherein the machine learning model is a multilayer perceptron, and rendering the digital asset includes: generating an input vector that corresponds to the intersection point based on the triplane representation; and evaluating the input vector using the multilayer perceptron to determine the color value (Section 2.3 teaches using a color prediction multilayer perceptron (MLP) to evaluate the color for a point. Thus, the machine learning model is a multilayer perceptron. It generates the 3D location of the intersection point and the interpolated triplane features as an output from an encoder and passes it into the MLP. Thus, the 3D location of the intersection point and triplane features can be considered the input vector).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu with the multilayer perceptron to determine a color value using the triplane representation taught by Gupta in order to allow for quick generation of textured 3D objects (Gupta Abstract).
19. Regarding claim 9, Sheppard in view of Yu and Gupta teaches the limitations of claim 1.
Sheppard further teaches the method wherein the at least one digital scene element includes one or more of a light source or an additional digital asset (Paragraphs 11 and 93 teach the virtual environment comprises of a plurality of objects. Thus, the one digital scene element in the digital scene is an additional digital asset).
20. Regarding claim 10, claim 10 is the system claim (Sheppard Paragraph 51 teaches a tangible non-transient computer readable medium which is a memory and a computer, or processor, that executes instructions) of method claim 1 and is accordingly rejected using substantially similar rationale as to that which is set for with respect to claim 1.
21. Regarding claim 18, claim 18 is the non-transitory computer-readable medium claim (Sheppard Paragraph 51 teaches a tangible non-transient computer readable medium which is a memory and a computer, or processor, that executes instructions) of method claim 1 and is accordingly rejected using substantially similar rationale as to that which is set for with respect to claim 1.
22. Regarding claim 22, Sheppard in view of Yu and Gupta teaches the limitations of claim 1. However, Sheppard is not relied upon for the below claim language: the method further comprising: 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 a ground truth radiance value; and generating the compressed representation by jointly training the triplane representation and the machine learning model using the training dataset, the jointly training including: predicting, using the compressed representation, an outgoing radiance value based on the respective training sample; and adjusting weights of the machine learning model and updating parameters of the triplane representation based on a comparison of the ground truth radiance value and the outgoing radiance value.
Yu teaches the method further comprising: generating, by the processing device, a training dataset that includes training images depicting the digital asset (Section 4.1 ‘Real image capture setup’ and ‘Synthetic Images’ sections teach generating a training dataset by capturing images of the object from different viewpoints),
the training dataset including one or more training samples per pixel that depicts the digital asset within the training images (Section 4.1 ‘Real image capture setup’ and ‘Synthetic Images’ sections teach generating a training dataset by capturing images of the object from different viewpoints. The different images depicting the digital asset at different viewpoints teach one or more training samples per pixel since the intersection point or pixel of the digital asset can be present in multiple different training images; Page 18, ‘Supplementary Material’ Section B teaches “For synthetic datasets, we sample Nc = 64 coarse samples and Nf = 128 fine samples per ray” This teaches there are multiple training samples),
a respective training sample of a respective pixel including a ground truth radiance value (Section 3.4 teaches a loss function using the training images which has the variable C(r). C(r) teaches the ground truth pixel color for the respective pixel that the ray or viewpoint intersects with);
and generating the compressed representation by jointly training (Section 3.4 and Equation 8 teaches a loss function that trains the neural model based on the predicted outgoing radiance value calculated by L_coarse and L_fine and the actual ground truth radiance or pixel color C(r). Section 3.4 teaches the loss function is used for the OSF learning which means the parameters in the neural model are updated which includes the machine learning parameters, taught in Section 3.5, in the OSF);.
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of rendering the digital asset in a scene taught by Sheppard with training the machine learning model using the training dataset taught by Yu in order to allow for free-viewpoint relight and scene composition of both opaque and translucent objects (Yu Section 1 Paragraph 8).
However, Sheppard and Yu are not relied upon for the below claim language: jointly training the triplane representation and the machine learning model using the training dataset; the jointly training including adjusting weights of the machine learning model and updating parameters of the triplane representation based on a comparison of the ground truth radiance value and the outgoing radiance value.
Gupta teaches jointly training the triplane representation and the machine learning model using the training dataset; the jointly training including adjusting weights of the machine learning model and updating parameters of the triplane representation based on a comparison of the ground truth radiance value and the outgoing radiance value (Section 2.2 teaches the model consists of the triplane representation and machine learning models like the PointNet, UNet, and MLP. Gupta teaches training the model and updating parameters of the triplane representation by using reconstruction loss and KL divergence loss which ensures the triplane feature distribution is close to a gaussian prior; Section 2.3 also teaches training the model through a loss function which compares the output color to the ground truth surface color. The output color teaches the outgoing radiance value and the ground truth surface color teaches the ground truth radiance. The loss function indicates the weights and parameters of the machine learning models are updated in training).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu with jointly training the triplane representation and machine learning model taught by Gupta in order to allow for quick generation of textured 3D objects (Gupta Abstract).
23. Regarding claim 23, Sheppard in view of Yu and Gupta teaches the limitations of claim 22. However, Sheppard is not relied upon for the below claim language: the method wherein the respective training sample includes a view direction, a shading direction representing an intersection point at which a primary ray traced from a virtual camera intersects with the three- dimensional geometry, a light direction to a light source in the digital scene, a visibility term that represents self-occlusion properties of the digital asset, and one or more surface properties of the digital asset.
Yu teaches the method wherein the respective training sample includes a view direction (Section 4.1 ‘Real image capture setup’ and ‘Synthetic images’ sections teach the training images are images from random viewpoints. This teaches the training sample includes a view direction. The Applicant does not define training sample so the training sample can be interpreted broadly to just be the training image collection),
a shading direction representing an intersection point at which a primary ray traced from a virtual camera intersects with the three- dimensional geometry (Section 3.4 teaches C(r) denotes the ground truth pixel color for ray ‘r’. Ray ‘r’ teaches the shading direction and represents an intersection point at which a primary ray traced from a virtual camera intersects with the three-dimensional geometry),
a light direction to a light source in the digital scene (Section 4.1 ‘Real image capture setup’ and ‘Synthetic images’ sections teach the training images are images with random light directions. This teaches the training images and samples include a light direction to a light source),
a visibility term that represents self-occlusion properties of the digital asset (Section 3.4 teaches C(r) denotes the ground truth pixel color for ray ‘r’. Ray ‘r’ is taught to be r=o-tw_out. The variable ‘t’ in is taught in Section 3.1 to represent the transmittance distribution along the ray. Section 3.6 ‘Direct lighting’ subsection teaches that the transmittance “can be seen as a soft measure for visibility” and thus can be used to solve the shadows or self-occlusion properties. Thus, the training sample that provided C(r) includes a visibility term representing self-occlusion properties; Figure 7 teaches the ground truth training image of an airplane as self-occlusion properties),
and one or more surface properties of the digital asset (Section 4.1 teaches the training dataset can includes translucent objects and opaque objects. Thus, the training images and samples include the digital asset’s surface property of being opaque or translucent).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of rendering the digital asset in a scene taught by Sheppard with the training sample taught by Yu in order to allow for free-viewpoint relight and scene composition of both opaque and translucent objects (Yu Section 1 Paragraph 8).
24. Claim(s) 3 and 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, as applied to claim 1 and 10 above, and further in view of Zheng et al. (“A Self-Occlusion Aware Lighting Model for Real-Time Dynamic Reconstruction”), hereinafter referred to as Zheng.
25. Regarding claim 3, Sheppard in view of Yu and Gupta teaches the limitations of claim 1. However, Sheppard, Yu, and Gupta are not relied upon for the below claim language: the method wherein the one or more lighting effects are based in part on a visibility term included in the compressed representation that represents self-occlusion properties of the digital asset.
Zheng teaches the method wherein the one or more lighting effects are based in part on a visibility term included in the compressed representation that represents self-occlusion properties of the digital asset (Section 3.2 and Equation 6 teach that the radiance for a surface point ‘n’ has a visibility term V(w) that represents the self-occlusion properties of the object. ‘w’ represents the direction of the light source).
Sheppard, Yu, Gupta, and Zheng are considered analogous to the claimed invention as because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Gupta and Yu with the visibility term taught by Zheng in order to render objects in varying lighting conditions and avoid a low quality appearance (Zheng Abstract).
26. Regarding claim 12, Sheppard in view of Yu and Gupta teaches the limitations of claim 10. However, Sheppard, Yu, and Gupta are not relied upon for the below claim language: the system wherein the one or more lighting effects are based in part on a visibility term included in the compressed representation that indicates whether portions of the digital asset are occluded by other portions of the digital asset.
Zheng teaches the system wherein the one or more lighting effects are based in part on a visibility term included in the compressed representation that indicates whether portions of the digital asset are occluded by other portions of the digital asset. (Section 3.2 and Equation 6 teach that the radiance for a surface point ‘n’ has a visibility term V(w) that represents the self-occlusion properties of the object. ‘w’ represents the direction of the light source. Self-occlusion means that a portion of the digital asset is occluded by another portion of the asset).
Sheppard, Yu, Gupta, and Zheng are considered analogous to the claimed invention as because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the system of including a digital asset into a digital scene taught by Sheppard in view of Yu and Gupta with the visibility term taught by Zheng in order to render objects in varying lighting conditions and avoid a low quality appearance (Zheng Abstract).
27. Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, and Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, as applied to claim 5, and further in view of Talegaonkar et al. (U.S. Patent Application Publication No. 2025/0131647 A1), hereinafter referred to as Talegaonkar.
Regarding claim 6, Sheppard in view of Yu and Gupta teaches the limitations of claim 5. However, Sheppard, Yu, and Gupta are not relied upon for the below claim language: the method wherein the input vector is a concatenation of a feature vector extracted from the triplane representation and a property vector that defines one or more properties of the digital asset at the intersection point.
Talegaonkar teaches the method wherein the input vector is a concatenation of a feature vector extracted from the triplane representation and a property vector that defines one or more properties of the digital asset at the intersection point (Paragraph 90 and Figure 6 teach concatenating triplane features and general features in step 614 before inputting it into the MLP 618. The general features concatenated with the triplane features can be considered the property vector and the triplane features can be considered the feature vector).
Sheppard, Yu, Gupta, and Talegaonkar are considered analogous to the claimed invention as because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu and Gupta with the concatenation taught by Talegaonkar in order to generate dynamic and detailed 3D models with MLPs (Talegaonkar Paragraph 35).
28. Claim(s) 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, and Talegaonkar et al. (U.S. Patent Application Publication No. 2025/0131647 A1), hereinafter referred to as Talegaonkar, as applied to claim 6 above, and further in view of Pfister et al. (U.S. Patent Application Publication No. 2006/0028474 A1), hereinafter referred to as Pfister, and Zheng et al. (“A Self-Occlusion Aware Lighting Model for Real-Time Dynamic Reconstruction”), hereinafter referred to as Zheng.
Regarding claim 8, Sheppard in view of Yu, Gupta, and Talegaonkar teaches the limitations of claim 6. However, Sheppard, Yu, Gupta, and Talegaonkar are not relied upon for the below claim language: the method wherein the digital asset is a surface-based digital asset and the property vector includes data to represent a view direction, a light direction, a normal, and a visibility term particular to the intersection point.
Pfister teaches the method wherein the digital asset is a surface-based digital asset (Paragraph 23 teaches using surface reflectance fields of objects, which means the object is a surface-based digital asset; Paragraph 46 and 51 teaches the objects, in which the surface reflectance field is calculated for, have surface normals. The Applicant only defines a “surface-based digital asset” to just include one or more surfaces in Paragraph 49 of the Applicant’s Specification. A surface normal indicates a surface) and the property vector includes data to represent a view direction, a light direction, a normal, (Paragraph 42 and Equation 2 teach the look-up function that maps to a vector has data on the view direction ‘v’ and light direction ‘l’ that is particular to the intersection point ‘p’. Paragraph 49 also teaches the directions l and v are unit vectors. Paragraph 51 also teaches the look up function that maps to a vector has data representing the model normal which approximates the surface normal particular to the intersection point ‘p’).
Sheppard, Yu, Gupta, Talegaonkar, and Pfister are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu, Gupta, and Talegaonkar with the property vector taught by Pfister in order to render deformed models and reduce the amount of data used when lighting an object under different lightings (Pfister Abstract and Paragraph 18).
However, Sheppard, Yu, Gupta, Talegaonkar, and Pfister are not relied upon for the below claim language: the property vector includes data to represent a visibility term particular to the intersection point.
Zheng teaches the property vector includes data to represent a light direction, a normal, and a visibility term particular to the intersection point (Section 3.2 and Equation 6 teach that the radiance for a surface normal for a particular intersection point ‘n’ has a visibility term V(w) that represents the self-occlusion properties of the object. ‘w’ represents the direction of the light source; Section 4.1 teaches ‘n’ represents the surface normal for a particular intersection point).
Sheppard, Yu, Gupta, Talegaonkar, Pfister, and Zheng are considered analogous to the claimed invention because all are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the method of including a digital asset into a digital scene taught by Sheppard in view of Yu, Gupta, Talegaonkar, and Pfister with the visibility term taught by Zheng in order to render objects in varying lighting conditions and avoid a low quality appearance (Zheng Abstract).
29. Claim(s) 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, and Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, and Zheng et al. (“A Self-Occlusion Aware Lighting Model for Real-Time Dynamic Reconstruction”), hereinafter referred to as Zheng, as applied to claim 12 above, and further in view of Pharr et al. (“Physically Based Rendering: From Theory to Implementation”), hereinafter referred to as Pharr.
Regarding claim 13, Sheppard in view of Yu, Gupta, and Zheng teaches the limitations of claim 12. However, Sheppard, Yu, Gupta, and Zheng are not relied upon for the below claim language: the system the operations further comprising computing the visibility term by tracing an occlusion ray from a first point on the compressed representation to a light source included in the digital scene and determining whether the occlusion ray intersects a second point of the compressed representation.
Pharr teaches the system the operations further comprising computing the visibility term by tracing an occlusion ray from a first point on the compressed representation to a light source included in the digital scene and determining whether the occlusion ray intersects a second point of the compressed representation (Chapter 1, Section 1.2 Page 5 teaches that visibility is determined by tracing a ray from the surface to the light. If the ray is uninterrupted then it is not occluded. This inherently includes determining if the ray intersected a second point).
Sheppard, Yu, Gupta, Zheng, and Pharr are considered analogous to the claimed invention as because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the system of including a digital asset into a digital scene taught by Sheppard in view of Yu, Gupta, and Zheng with the computation of the visibility term taught by Pharr in order to achieve photorealistic rendering (Chapter 1, Section 1.2, Paragraph 1).
30. Claim(s) 14-16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, as applied to claim 10 above, and further in view of Rainer et al. (“Neural Precomputed Radiance Transfer”), hereinafter referred to as Rainer.
31. Regarding claim 14, Sheppard in view of Yu and Gupta teaches the limitations of claim 10. However, Sheppard, Yu, and Gupta are not relied upon for the below claim language: the system with operations further comprising generating a training dataset that includes training images depicting the digital asset viewed from a plurality of different locations within the digital scene, and training the neural model using the training dataset.
Rainer teaches the system with operations further comprising generating a training dataset that includes training images depicting the digital asset viewed from a plurality of different locations within the digital scene, and training the neural model using the training dataset (Section 6, ‘Data Generation’ subsection teaches randomly placing the cameras and viewing direction to obtain training images depicting the scene and thus objects in the scene from various directions; Section 6 Paragraph 1 teaches using the data generated or training images to train the neural network; Section 2 Paragraph 1 teaches the methods taught by Rainer are run on synthetic data where the scene has been constructed by an artist. Thus, the scene can be considered a digital scene).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Rainer is considered analogous to the claimed invention because both are in the same field of relighting. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the system of including a digital asset into a digital scene taught by Sheppard in view of Yu and Gupta with the training dataset taught by Rainer in order to make sure all renderings contribute valuable information to the learning (Rainer Section 6, Paragraph 5).
32. Regarding claim 15, Sheppard in view of Yu, Gupta, and Rainer teaches the limitations of claim 14. However, Sheppard is not relied upon for the below claim language: the system wherein the machine learning model is a multilayer perceptron, and generating the compressed representation includes utilizing the training dataset to adjust weights of the multi-layer perceptron, and update parameters of the triplane representation to incorporate information learned from the training dataset into the triplane representation.
Gupta teaches the system wherein the machine learning model is a multilayer perceptron, and generating the compressed representation includes utilizing the training dataset to adjust weights of the multi-layer perceptron (Section 2.3 teaches the MLP is trained by calculating a loss comparing the output to the ground truth surface colors. Thus, the weights or parameters of the MLP is adjusted and trained. The MLP is the machine learning model), and update parameters of the triplane representation to incorporate information learned from the training dataset into the triplane representation (Section 2.2 paragraph 4 teaches the VAE which consists of creating the triplane representation is trained and updated using loss functions. The loss functions are information learned from the training dataset. Thus, the triplane representation is updated by incorporating information learned from the training dataset; Section 3.2 teaches the VAE is trained using multiple camera angles per object which is the training dataset of images from different viewpoint locations).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Rainer is considered analogous to the claimed invention because both are in the same field of relighting. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the system of including a digital asset into a digital scene taught by Sheppard in view of Yu and Rainer with the updating of the triplane representation taught by Gupta in order to allow for quick generation of textured 3D objects (Gupta Abstract).
33. Regarding claim 16, Sheppard in view of Yu, Gupta, and Rainer teach the limitations of claim 14. However, Sheppard is not relied upon for the below claim language: wherein the training dataset includes training samples for each pixel of the training images, the training samples including a visibility term that represents self-occlusion properties of the digital asset.
Yu teaches wherein the training dataset includes training samples for each pixel of the training images, the training samples including a visibility term that represents self-occlusion properties of the digital asset (Section 3.4 teaches C(r) denotes the ground truth pixel color for ray ‘r’. Ray ‘r’ is taught to be r=o-tw_out. The variable ‘t’ in is taught in Section 3.1 to represent the transmittance distribution along the ray. Section 3.6 ‘Direct lighting’ subsection teaches that the transmittance “can be seen as a soft measure for visibility” and thus can be used to solve the shadows or self-occlusion properties. Thus, the training sample that provided C(r) includes a visibility term representing self-occlusion properties).
Sheppard, Yu, and Gupta are considered analogous to the claimed invention because both are in the same field of rendering 3D objects. Rainer is considered analogous to the claimed invention because both are in the same field of relighting. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the system of rendering the digital asset in a scene taught by Sheppard with the training sample taught by Yu in order to allow for free-viewpoint relight and scene composition of both opaque and translucent objects (Yu Section 1 Paragraph 8).
34. Claim(s) 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sheppard et al. (U.S. Patent Application Publication No. 2016/0078640 A1), hereinafter referred to as Sheppard, in view of Yu et al. (“Learning Object-Centric Neural Scattering Functions for Free-Viewpoint Relighting and Scene Composition”), hereinafter referred to as Yu, and Gupta et al. (“3DGen: Triplane Latent Diffusion for Textured Mesh Generation”), hereinafter referred to as Gupta, as applied to claim 18 above, and further in view of Everitt et al. (U.S. Patent Publication No. 8,102,393 B1), hereinafter referred to as Everitt.
Regarding claim 20, Sheppard in view of Yu and Gupta teach the limitations of claim 18. However, Sheppard, Yu, and Gupta are not relied upon for the below claim language: the non-transitory computer-readable medium wherein the rendering the digital asset includes using one or more of a path tracer or a rasterizer-based renderer.
Everitt teaches the non-transitory computer-readable medium wherein the rendering the digital asset includes using one or more of a path tracer or a rasterizer-based renderer (Column 5 Line 56 – Column 6 Line 5 teaches a rasterizer that renders a digital asset).
Sheppard, Yu, Gupta, and Everitt are considered analogous to the claimed invention as because both are in the same field of rendering 3D objects. Thus, it would have been obvious to a person holding ordinary skill in the art before the effective filing date to modify the non-transitory computer-readable medium of including a digital asset into a digital scene taught by Sheppard in view of Yu and Gupta with the rasterizer taught by Everitt in order to render a 3D object by optimizing hardware resources and performance (Everitt Abstract).
Allowable Subject Matter
35. Claims 7 and 24 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), 2nd paragraph, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter:
In claim 7, the prior art fails to teach the method wherein the digital asset is a fiber- based digital asset that includes one or more fiber primitives and the property vector includes data to represent a view direction, a light direction, a tangent, a cross- section offset, and a visibility term particular to the intersection point.
In claim 24, the prior art fails to teach the non-transitory computer-readable medium rendering the digital asset includes: generating a feature vector by querying the triplane representation at a coordinate corresponding to an intersection point of a ray with the three-dimensional geometry; generating a property vector that includes a view direction and a light direction particular to the intersection point; generating an input vector by concatenating the feature vector and the property vector; and generating, by the machine learning model, a radiance value for the intersection point by processing the input vector based on the precomputed light transport.
Conclusion
36. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
- Harviainen (U.S. Patent Application Publication No. 2024/0087217 A1) teaches adding and rendering a 3D asset to a 3D scene and lighting the scene.
- Bi et al. (U.S. Patent Application Publication No. 2022/0335636 A1) teaches scene reconstruction with view synthesis and relighting of 3D objects.
- Lee et al. (U.S. Patent Application Publication No. 2024/0177414 A1) teaches rendering a 3D scene with tri-plane representations.
37. Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHRISTINE Y AHN whose telephone number is (571)272-0672. The examiner can normally be reached M-F 9-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alicia Harrington can be reached at (571)272-2330. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/CHRISTINE YERA AHN/Examiner, Art Unit 2615
/ALICIA M HARRINGTON/Supervisory Patent Examiner, Art Unit 2615
Prosecution Insights