DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of the Claims
Claims 1-20 are currently pending in the present application, with claims 1, 11, and 20 being independent.
Response to Amendments / Arguments
Applicant's arguments filed 04/14/2026 have been fully considered but they are not persuasive.
Applicant argues: Melas-Kyriazi et al. "Pc2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction." In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 12923-12932. 2023, hereinafter referred to as “Melas-Kyriazi”, does not disclose “generating a complex gaussian distribution…”and states that Gaussian noise used as an input to a diffusion or denoising process is distinct from a Gaussian distribution generated to represent an object.
Examiner replies: that Melas-Kyriazi expressly discloses “formulate 3D reconstruction as conditional generate: the target distribution q(X_0 | I, V) where I is an input image and V is the corresponding camera view” (Section 3.3, Par. 1), and further discloses “at inference time, a random point cloud XT ∼ N(0,I3N) is sampled…Section 3.4, Par. 2; Xt is the partially-noised point cloud” (Section 3.2, Par. 3). Under broadest reasonable interpretation, “generating a complex gaussian distribution…” reads on defining or establishing the conditional target distribution controlling generation. Melas-Kyriazi is operating in a probabilistic generative framework where the object point cloud is generated by sampling from a target conditional distribution, implemented via Gaussian-noise initialization and iterative denoising, where the conditional generative distribution “q(X_0 | I, V)” that governs sampling (Section 3.3) reads on the claimed “complex Gaussian distribution”. Melas-Kyriazi does not merely “use noise”, it defines and samples from the conditional generative distribution that produces the reconstructed object point cloud, and the claim does not require an explicit, standalone distribution artifact or separately output explicit Gaussian parameter set. Therefore, under broadest reasonable interpretation, Melas-Kyriazi discloses the claimed “generating a complex gaussian distribution…”.
Applicant argues: Melas-Kyriazi et al. "Pc2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction." In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 12923-12932. 2023, hereinafter referred to as “Melas-Kyriazi”, does not disclose “specifications for a physical component” and further asserts that a camera view describing imaging perspectives used to capture images of a component does not correspond to specifications of a physical component such as geometric properties for the physical component.
Examiner replies: that the claim limitations do not limit the “set of specification” to intrinsic geometric properties such as desired thickness, desired width, a desired height, a desired curvature, etc., for the physical component. Under broadest reasonable interpretation, “specifications for a physical component” reasonably includes parameters used by the system to specify or define the component’s generation conditions. Melas-Kyriazi expressly discloses reconstruction on (I, V), where V is the camera view/pose corresponding to the component’s observation and is used as an input to the conditional generation formulation (Section 3.3, Par. 1). Thus, the camera view V is an explicit specification used in the generation of the 3D representation of the component and therefore reads on the “set of specifications” recited in claim 1.
Applicant argues: Yang et al. "Pointflow: 3d point cloud generation with continuous normalizing flows." In Proceedings of the IEEE/CVF international conference on computer vision, pp. 4541-4550. 2019, hereinafter referred to as “Yang”, does not disclose “generating a complex Gaussian distribution…based on meshes of a physical component…” and asserts that Yang merely describes learning a latent distribution over shape representations rather than a Gaussian distribution generated based on meshes of a physical component.
Examiner replies: Under broadest reasonable interpretation, the limitation is satisfied where mesh-based geometry is used as the basis for the probabilistic model/distribution that controls generation. Yang expressly discloses that the point clouds used for the shapes are obtained by sampling points uniformly from mesh surface (Section 6.2, Par. 1), and further teaches a probabilistic generative framework with Gaussian priors that are transformed via a continuous normalizing flow conditioned on a shape representation z (Section 5.1-5.2), and therefore models a learnable distribution over shapes derived from the mesh-based training data. Thus, Yang teaches generating a distribution over shapes (mesh-derived geometry), using Gaussian priors and transformations, which reads on the claimed limitation of “generating a complex Gaussian distribution…based on meshes of a physical component”. Further, Melas-Kyriazi discloses the image and view-conditioned target distribution for reconstruction (Section 3.3), Yang discloses the mesh-derived shape prior distribution over shapes from mesh-sampled point cloud (Section 5 and 6.2), where Yang supplies the mesh-based conditioning used in the combined system.
In response to applicant’s argument that there is no teaching, suggestion, or motivation to combine the references, the examiner recognizes that obviousness may be established by combining or modifying the teachings of the prior art to produce the claimed invention where there is some teaching, suggestion, or motivation to do so found either in the references themselves or in the knowledge generally available to one of ordinary skill in the art. See In re Fine, 837 F.2d 1071, 5 USPQ2d 1596 (Fed. Cir. 1988), In re Jones, 958 F.2d 347, 21 USPQ2d 1941 (Fed. Cir. 1992), and KSR International Co. v. Teleflex, Inc., 550 U.S. 398, 82 USPQ2d 1385 (2007). In this case, Melas-Kyriazi and Yang are in the same technical field of probabilistic generative modeling for 3D point sets, and both start from Gaussian priors or noise and use learned transformation to obtain realistic point clouds/shapes. A person of ordinary skill in the art would have been motivated to incorporate Yang’s mesh-derived shape prior conditioning into Melas-Kyriazi’s reconstruction to provide additional component-specific geometric constraints that guide sampling/denoising, yielding predictable benefits of improved reconstruction consistency and shape fidelity. The examiner respectfully asserts that this is not hindsight, it is a routine design choice in generative modeling to strengthen conditioning by adding an additional conditioning factor (such as geometric properties) to an existing conditional generator (conditioning on image and views). Melas-Kyriazi already uses conditioning inputs to guide denoising, and Yang provides a learned shape representation from mesh-derived data. Integrating an additional conditioning vector into a conditional denoiser is a predictable modification with predictable results of improved shape generation.
Claim Rejections - 35 USC § 103
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.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Melas-Kyriazi et al. "Pc2: Projection-conditioned point cloud diffusion for single-image 3d reconstruction." In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 12923-12932. 2023, hereinafter referred to as “Melas-Kyriazi”, in view of Yang et al. "Pointflow: 3d point cloud generation with continuous normalizing flows." In Proceedings of the IEEE/CVF international conference on computer vision, pp. 4541-4550. 2019. hereinafter referred to as “Yang”.
Regarding claim 1, Melas-Kyriazi discloses a method, comprising:
generating a complex Gaussian distribution based on a set of specifications for a physical component, a set of images of the physical component (Section 3.3, Par. 1; formulate 3D reconstruction as conditional generate: the target distribution q(X_0 | I, V) where I is an input image and V is the corresponding camera view. Section 3.4, Par. 2; let I ∈ RH×W×C be the feature volume produced by our 2D image model applied to our input image, where C is the number of feature channels. sθ is now a function R(3+C)N → R3N which predicts the noise ϵ from the augmented point cloud X+t = [Xt,Xprojt]. The projected features Xprojt are given by Xprojt =PVI(I,Xt), where PVI is the projection function from camera view VI, I is the input image. Fig. 2; At each step in the diffusion process, we project image features onto the partially-denoised point cloud from the given camera pose, augmenting each point with a set of neural features. Fig. 3; input images. Examiner's note: set of specifications is the corresponding camera view, set of images is input images and its corresponding image features, and the conditional distribution is the complex Gaussian distribution. See more in examiner’s response to arguments),
obtaining a randomly generated point cloud (Section 3.2, Par. 3; at inference time, a random point cloud XT ∼ N(0,I3N) is sampled…Section 3.4, Par. 2; Xt is the partially-noised point cloud),
and generating a component point cloud based on the complex Gaussian distribution, the randomly generated point cloud, and a diffusion denoising model, wherein the component point cloud represents the physical component (Section 3.2; diffusion model…This network denoises the positions of a set of points from a spherical Gaussian ball into a recognizable object…a random point cloud XT ∼ N(0,I3N) is sampled from a 3N-dimensional Gaussian and the reverse diffusion process is run to produce a sample X0. Section 3.4 PC^2: Projection-Conditinal Diffusion Models; …a point cloud augmented with projection conditioning…Examiner's note: output is the reconstructed 3D object point cloud conditioned on the input image and camera view).
Melas-Kyriazi does not disclose generating a complex Gaussian distribution based on
In the same art of 3D point cloud generation, Yang discloses generating a complex Gaussian distribution based on a set of meshes for the physical component (Section 3; …Each shape Xi is itself a sample from a distribution over shapes Q(X) that captures what shapes in this category look like…Section 5; a prior Pψ(z) over shape representations. Section 6.2, Par. 1; The point clouds are obtained by sampling points uniformly from the mesh surface).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Yang’s mesh-based conditioning into Melas-Kyriazi’s conditioned point-cloud diffusion system. Doing so would provide additional component-specific information to guide diffusion denoising, yielding predictable results in improved reconstruction consistency and shape fidelity when generating a 3D component.
Examiner’s note: As noted above in examiner’s response to arguments, Melas-Kyriazi and Yang are in the same technical field of probabilistic generative modeling for 3D point sets, and both start from Gaussian priors or noise and use learned transformation to obtain realistic point clouds/shapes. A person of ordinary skill in the art would have been motivated to incorporate Yang’s mesh-derived shape prior conditioning into Melas-Kyriazi’s reconstruction to provide additional component-specific geometric constraints that guide sampling/denoising, yielding predictable benefits of improved reconstruction consistency and shape fidelity. The examiner respectfully asserts that this is not hindsight, it is a routine design choice in generative modeling to strengthen conditioning by adding an additional conditioning factor (such as geometric properties) to an existing conditional generator (conditioning on image and views). Melas-Kyriazi already uses conditioning inputs to guide denoising, and Yang provides a learned shape representation from mesh-derived data. Integrating an additional conditioning vector into a conditional denoiser is a predictable modification with predictable results of improved shape generation.
Regarding claim 2, Melas-Kyriazi in view of Yang discloses the method of claim 1, and further discloses wherein obtaining the complex Gaussian distribution comprises:
generating a first vector based on the set of specifications (Melas-Kyriazi Section 3.3, Par. 1; formulate 3D reconstruction as conditional generate: the target distribution q(X_0 | I, V) where I is an input image and V is the corresponding camera view. Section 3.4, Par. 2; …PVI is the projection function from camera view VI);
generating a second vector based on the set of images (Melas-Kyriazi Section 3.3, Par. 1; formulate 3D reconstruction as conditional generate: the target distribution q(X_0 | I, V) where I is an input image and V is the corresponding camera view. Section 3.4, Par. 2; …I is the input image);
and generating a combined vector based on the first vector, the second vector (Melas-Kyriazi Section 3.4, Par. 2; augmented point cloud X+t = [Xt,Xprojt])
Melas-Kyriazi does not disclose generating a third vector based on the set of meshes, and generating a combined vector
In the same art of 3D point cloud generation, Yang discloses generating a third vector based on the set of meshes (Section 5; a prior Pψ(z) over shape representations. Section 6.2, Par. 1; The point clouds are obtained by sampling points uniformly from the mesh surface), and generating a combined vector based on the third vector (Section 5-5.1; point cloud into a shape representation z…a point x in the point set X is the result of transforming some point y(t0) in the prior distribution P(y)=N(0,I) using a CNF conditioned on z)
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to Yang’s conditioning vector z of a shape representation into Melas-Kyriazi’s conditioning input concatenated with the existing conditional inputs/feature representations of an input image and camera view. Melas-Kyriazi explicitly combines point and projected image features into an augmented representation, and Yang explicitly conditions its generative mapping on a learned shape representation z, therefore combining multiple conditioning vectors is a routine design choice in conditional generative models to incorporate complementary sources of information into a single conditioning input used by the diffusion model, yielding predictable results in improved control over the output distribution.
Regarding claim 3, Melas-Kyriazi in view of Yang discloses the method of claim 2, and further discloses wherein obtaining the complex Gaussian distribution further comprises:
generating the complex Gaussian distribution based on the combined vector (Yang Section 5-5.1; point cloud into a shape representation z…a point x in the point set X is the result of transforming some point y(t0) in the prior distribution P(y)=N(0,I using a CNF conditioned on z), a simple Gaussian distribution (Yang Section 4.2, Par. 1; a latent variable z with a prior distribution Pψ(z). Section 5.2; transforming a simple Gaussian P(w) = N(0, 1) with a CNF), and a reverse conditional normalizing flow model (Yang Section 5.1; gθ defines the continuous-time dynamics of the flow Gθ conditioned on z. ..the inverse of Gθ is given by…Fig. 2).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Yang’s conditional continuous normalizing flow and its inverse form to transform a simple Gaussian into a more conditioned distribution into Melas-Kyriazi’s diffusion-based point-cloud generation system. It is common for diffusion models to start from a simple Gaussian, and Yang explicitly teaches that a standard Gaussian prior is too restrictive for capturing the variability of real 3D point-set distributions (Section 5.1; Although it is possible to use a simple Gaussian prior over shape representations, it has been shown that a restricted prior tends to limit the performance of VAEs), therefore, the combination provides a more expressive, condition-controlled latent distribution, yielding predictable results in improved stability and control when generating 3D point cloud from conditioned inputs.
Regarding claim 4, Melas-Kyriazi in view of Yang discloses the method of claim 3, and further discloses wherein the reverse conditional normalizing flow model transforms the simple Gaussian distribution to the complex Gaussian distribution based on the combined vector (Yang Section 5.2; transforming a simple Gaussian P(w) = N(0, 1) with a CNF…the inverse of Fψ is given by…).
Melas-Kyriazi and Yang are combined for the reason set forth above with respect to claim 3.
Regarding claim 5, Melas-Kyriazi in view of Yang discloses the method of claim 3, and further discloses wherein the combined vector indicates one or more conditions for transforming the simple Gaussian distribution to the complex Gaussian distribution (Yang Section 5.1; a point x in the point set X is the result of transforming some point y(t0) in the prior distribution P(y)=N(0,I using a CNF conditioned on z).
Melas-Kyriazi and Yang are combined for the reason set forth above with respect to claim 3.
Regarding claim 6, Melas-Kyriazi in view of Yang discloses the method of claim 3, and further discloses wherein the simple Gaussian distribution is randomly generated (Yang Section 5.1-5.2; prior distribution P(y)=N(0,I)…simple Gaussian P(w)=N(0,I). Footnote 1; randomly sampling points from the underlying distribution).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Yang’s randomly generated simple Gaussian distribution into Melas-Kyriazi’s conditioned point-cloud diffusion system. Doing so is a standard approach, allowing efficient distribution for sampling in both diffusion and flow frameworks, yielding predictable results in enabling diversity of outputs and stable numerical behavior.
Regarding claim 7, Melas-Kyriazi in view of Yang discloses the method of claim 3, and further discloses wherein generating the component point cloud comprises:
moving a set of points from the randomly generated point cloud based on the complex Gaussian distribution, wherein the complex Gaussian distribution indicates one or more points that can be moved (Yang Section 1, Par. 2; generating points for a given shape involved sampling points from a generic Gaussian prior, and then moving them according to this parameterized transformation to their new location in the target shape, as illustrated in Figure 1. Fig. 2; we first sample points from the 3-D Gaussian prior and then move them according to the CNF parameterized by z).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate generation as moving points from an initial randomly generated point cloud based on a complex Gaussian distribution. Using this iterative refinement technique provides a computationally efficient technique that yields predictable results in smoother optimization and sampling trajectories, improved stability, and provide stable refinement from noise to structure in diffusion and flow methods.
Regarding claim 8, Melas-Kyriazi in view of Yang discloses the method of claim 1, and further discloses wherein the set of specifications indicate one or more physical properties of the physical component (Melas-Kyriazi Section 3.4, Par. 2; The projected features Xprojt are given by Xprojt =PVI(I,Xt), where PVI is the projection function from camera view VI. Examiner's note: acquisition/view parameters used to represent the component)
Melas-Kyriazi and Yang are combined for the reason set forth above with respect to claim 1.
Regarding claim 9, Melas-Kyriazi in view of Yang discloses the method of claim 1, and further discloses wherein the set of images comprise one or more of 2-dimensional images and color images (Melas-Kyriazi Section 1, Par. 2; RGB image data. Section 3.4, Par. 2; let I ∈ RH×W×C be the feature volume produced by our 2D image model applied to our input image).
Melas-Kyriazi and Yang are combined for the reason set forth above with respect to claim 1.
Regarding claim 10, Melas-Kyriazi in view of Yang discloses the method of claim 1, and further discloses wherein the set of meshes indicate physical properties of one or more other components that may interact with the physical component (Yang Section 5-5.1; …shape representation z…).
Melas-Kyriazi and Yang are combined for the reason set forth above with respect to claim 1.
Regarding claim 11, claim 11 is the system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU) of method claim 1, and is accordingly rejected using substantially similar rationale as to that which is set for with respect to claim 1.
Regarding claim 12, claim 12 has similar limitations as of claim 2, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 2.
Regarding claim 13, claim 13 has similar limitations as of claim 3, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 3.
Regarding claim 14, claim 14 has similar limitations as of claim 4, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 4.
Regarding claim 15, claim 15 has similar limitations as of claim 5, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 5.
Regarding claim 16, claim 16 has similar limitations as of claim 6, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 6.
Regarding claim 17, claim 17 has similar limitations as of claim 7, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 7.
Regarding claim 18, claim 18 has similar limitations as of claim 8, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 8.
Regarding claim 19, claim 19 has similar limitations as of claim 9, except it is a system claim (Melas-Kyriazi Section 4, Implementation Details; …GPU), therefore it is rejected under the same rationale as claim 9.
Regarding claim 20, claim 20 is the CRM claim (Melas-Kyriazi Section 4, Implementation Details) of method claim 1, and is accordingly rejected using substantially similar rationale as to that which is set for with respect to claim 1.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JENNY NGAN TRAN whose telephone number is (571)272-6888. The examiner can normally be reached Mon-Thurs 8am-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.
/JENNY N TRAN/Examiner, Art Unit 2615
/ALICIA M HARRINGTON/Supervisory Patent Examiner, Art Unit 2615
Prosecution Insights