DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1-15 and 18-24 are presented for examination based on the amended claims in the application filed on July 31, 2026. Claims 16-17 have been cancelled by the applicant.
Claims 1, 9-11, 15, 19-22 are rejected under 35 U.S.C. 103 as being unpatentable over US 2019/0164055 A1 Ljung Larhed, et al. [herein “Ljung Larhed”] in view of Heimann, et a l. “3D Active Shape Models Using Gradient Descent Optimization of Description Length” Information Processing in Medical Imaging, pp. 566-577 (July 2005) [herein “Heimann”].
Claims 2-3, 5, 7, 12, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed and Heimann as applied to claims 1 and 11 above, and further in view of Bronstein et al. “Geometric Deep Learning: Going beyond Euclidean data” IEEE Signal Processing Magazine Volume: 34, Issue: 4 (July 2017), [herein “Bronstein”].
Claims 4, 6, and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed, Heimann, and Bronstein as applied to claims 2 and 12 above, and further in view of US Patent 9,922,432 B1 Risser.
Claims 8 and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed and Heimann as applied to claim 1, and further in view of US 2007/0055401 A1 Van Bael et al.
Claims 18 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed and Heimann as applied to claim 11 above, and further in view of US Patent 9,922,432 B1 Risser.
This action is made Non-Final.
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on July 31, 2026 has been entered.
Response to Amendment
The amendment filed July 31, 2026 has been entered. Claims 1-15 and 18-24 remain pending in the application. Applicant’s amendments to the Claims have overcome the 35 USC § 101 rejection previously set forth in the Final Office Action mailed July 31, 2026. In an interview with the Applicant’s representative, Gregory Suh (Registration No. 48,187) on August 19, 2026, the examiner recommended incorporating the subject matter of both claims 19 and 23 into the independent claims to overcome the 35 USC § 103 (further interview summary and response to arguments).
Information Disclosure Statement
The information disclosure statements (IDSs) submitted on February 27, 2026 and March 31, 2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Claim Objections
Claims 1-15 and 18-24 are objected to because of the following informality: recitations of elements with a previous recitation. For example, claim 1, “a direction” in Ln. 21, is improper because there has been a previous recitation of “a direction” in Ln. 16. For the purpose of examination, “a direction” in Ln. 21 will be interpreted as “the direction”. Similarly, the following are objected under similar rationale:
Claim 1, “a first sample point” in Ln. 22 should be “the first sample point”. Claims 11 and 20, having similar limitations of claim 1, are also objected.
Claims 11 and 20, having similar limitations of claim 1, are also objected.
All claims dependent on an objected base claim are objected based on their dependency.
Appropriate correction is required.
Claim 23 is objected to because of the following informality: recitations of elements with no previous recitations. For example, claim 23, “the UV domain” in Ln. 4, is improper because there has been no previous recitation of “the UV domain”. For the purpose of examination, “the UV domain” will be interpreted as “a the UV domain”.
Claim Rejections - 35 U.S.C. § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. § 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. § 102(b)(2)(C) for any potential 35 U.S.C. § 102(a)(2) prior art against the later invention.
Claims 1, 9-11, 15, 19-22 are rejected under 35 U.S.C. 103 as being unpatentable over US 2019/0164055 A1 Ljung Larhed, et al. [herein “Ljung Larhed”] in view of Heimann, et a l. “3D Active Shape Models Using Gradient Descent Optimization of Description Length” Information Processing in Medical Imaging, pp. 566-577 (July 2005) [herein “Heimann”].
As per Claim 1, Ljung Larhed teaches “A computer-implemented method for generating one or more visualizations of at least one geometric style gradient”. (Para. 21, “With the present disclosure, machine learning is used to find an n-dimensional identifier of a 3D object that captures particular features relevant to distinguish that object from other 3D objects. Objects that are similar generate similar feature values” [geometric style]. Para. 44, “The neural network 812 is adjusted so that the 0.43 and 0.48 output values 810 are brought closer together (i.e., a smaller difference) by causing an increase in the 0.43 value (output) and a decrease in the 0.48 value (output) for this particular feature. For example, the adjusted values may be passed in the reverse direction (backward propagating) to the neural network 812 to converge outputs to the same or similar value” [values brought closer to similar value to show similarity, i.e., a gradient]. Para. 28, “The neural network training system 200 in one example uses back propagation or other training techniques. The neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object ( e.g., 3D mesh) that captures particular features relevant to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects (that are not identical)” [using back propagation on the particular features to generate a n-dimensional identifier, e.g., generating a geometric style gradient]. Para. 59, “the method 1000 tunes the neural network to make the outputs similar for the similar training meshes” [e.g., a method for generating]. Para. 0063, “The computing apparatus 1102 may comprise an input/output controller 1118 configured to output information to one or more input devices 1120 and output devices 1122, for example a display” [e.g., a computer-implemented method for generating one or more visualizations of at least one geometric style gradient]. Further see Para 21, 28, 44, 59, and 63. The examiner has interpreted that a method that distinguishes an object from other 3D objects using particular features through the use of back propagation to find an n-dimensional identifier that is adjusted to increase the output value to converge to a similar value for finding similar objects and is outputted through a display device as a computer-implemented method for generating one or more visualizations of at least one geometric style gradient.)
Ljung Larhed also teaches “generating, via execution of a trained neural network, a first plurality of style signals based on a first three dimensional (3D) computer-aided design (CAD) object.” (Para. 21, “the neural network is trained to output relevant distinguishing features by using slightly modified input geometry meshes” [a first plurality of style signals]. Para 21, “machine learning is used to find an n-dimensional identifier of a 3D object that captures particular features relevant to distinguish that object from other 3D objects. Objects that are similar generate similar feature values” [generating, via execution of a trained neural network, a first plurality of style signals based on a first three dimensional (3D) computer-aided design (CAD) object]. Further see Para. 21. The examiner has interpreted the capturing particular relevant features of a 3D object to distinguish that object from other 3D objects using a neural network as generating, via execution of a trained neural network, a first plurality of style signals based on a first three dimensional (3D) computer-aided design (CAD) object.)
Ljung Larhed also teaches “generating, via execution of the trained neural network, a second plurality of style signals based on a second 3D CAD object.” (Para. 28, “the neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object {e.g., 3D mesh} that captures particular features relevant” [a second plurality of style signals style signals] “to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects” [generating, via execution of the trained neural network, a second plurality of style signals based on a second 3D CAD object]. Further see Para. 28. The examiner has interpreted the particular relevant features generating distinguishing feature values for other 3D objects using a neural network as generating, via execution of the trained neural network, a second plurality of style signals based on a second 3D CAD object.)
Ljung Larhed also teaches “computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric [for each position included in a first plurality of positions] associated with the first 3D CAD object to generate a first geometric style gradient.” (Para. 21, “With the present disclosure, machine learning is used to find an n-dimensional identifier of a 3D object that captures particular features relevant to distinguish that object from other 3D objects. Objects that are similar generate similar feature values” [geometric style and based on the first plurality of style signals and the second plurality of style signals]. Para. 45, “the search at 906 in one example includes identifying meshes having similar features defined by similar outputs, such as meshes having output values for the relevant properties within a predetermined variance of the values for the relevant features” [style comparison metrics] “of the unknown mesh 902 input into the neural network 900. The trained neural network 900 can be used on the unknown mesh 902 to produce the set of features 904 used to search a database of pre-existing 3D object feature values to identify any similar meshes.” Para. 44, “The neural network 812 is adjusted so that the 0.43 and 0.48 output values 810 are brought closer together (i.e., a smaller difference) by causing an increase in the 0.43 value (output) and a decrease in the 0.48 value (output) for this particular feature. For example, the adjusted values may be passed in the reverse direction (backward propagating) to the neural network 812 to converge outputs to the same or similar value” [values brought closer to similar value to show similarity, i.e., geometric style gradient]. Para. 28, “The neural network training system 200 in one example uses back propagation or other training techniques. The neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object ( e.g., 3D mesh) that captures particular features relevant to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects (that are not identical)” [using back propagation on the particular features to generate a n-dimensional identifier and different partial derivatives are the core of back propagation, e.g., computing a different partial derivative associated with the first 3D object to generate a first geometric style gradient]. On a further note, the examiner has interpreted the relevant properties [i.e., features] as styles and the output values of relevant features as the style comparison metrics. Further see Para 21, 28, 44-45, and 59. The examiner has interpreted that distinguishing an object from other 3D objects using particular features through the use of back propagation to find an n-dimensional identifier that is adjusted to increase the output value for the relevant feature properties to converge to a similar value for finding similar objects as computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric associated with the first 3D CAD object to generate a first geometric style gradient.)
Ljung Larhed also teaches “wherein the first geometric style gradient [includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point] that indicates a direction and magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object [at the first position]”. (Para. 28, “The neural network training system 200 in one example uses back propagation or other training techniques. The neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object ( e.g., 3D mesh) that captures particular features relevant to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects (that are not identical)” [using back propagation on the particular features to generate a n-dimensional identifier and different partial derivatives are the core of back propagation, e.g., computing the first geometric style gradient]. Para. 44, “The neural network 812 is also adjusted so that the 0.43 and 0.12 output values 810 are made farther apart (i.e. a greater difference) by causing an increase in the 0.43 value (output) and a decrease in the 0.12 value (output) for this particular feature” [values diverge further away, e.g., the first geometric style gradient that indicates a magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object]. “For example, the adjusted values may be passed in the reverse direction (backward propagating) to the neural network 812 to diverge outputs to the dissimilar values. The adjustment is performed for one or all of the output values 810 such that the neural network 812 is trained such that the output values 810 for the first and second meshes 800 and 802 can be identified as being similar and the outputs values 810 for the first and third meshes 800 and 806 can be identified as being dissimilar” [values diverge toward each other in reverse direction, e.g., changing the sign, the first geometric style gradient that indicates a geometric style dissimilarity]. Further see Para. 28 and 44. The examiner has interpreted that distinguishing an object from other 3D objects using particular features through the use of back propagation to find an n-dimensional identifier that is adjusted to increase the output value for the relevant feature properties to diverge away from a similar value in the reverse direction for finding objects that are dissimilar as wherein the first geometric style gradient that indicates a direction and magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object.)
Ljung Larhed does not specifically teach a “[computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric] for each position included in a first plurality of positions associated with the first 3D CAD object [to generate a first geometric style gradient] wherein the first plurality of positions corresponds to a first plurality of sample points for the first 3D CAD object]”, “[wherein the first geometric style gradient] includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point [that indicates a direction and magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object] at the first position”, “generating a first graphical element based on the first vector included in the first geometric style gradient, wherein the first graphical element illustrates a direction and magnitude in which to move a first sample point corresponding to the first position for modifying the first 3D CAD object to increase a geometric style similarity between the first 3D CAD object and the second 3D CAD object at the first position”, and “positioning the first graphical element relative to the first position on the first 3D CAD object within a graphical user interface to generate a first visualization of the first geometric style gradient.”
However, in the same field of endeavor namely creating geometric gradients, Heimann teaches “computing, [based on the first plurality of style signals and the second plurality of style signals,] a different partial derivative of a style comparison metric for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient, wherein the first plurality of positions corresponds to first plurality of sample points for the first 3D CAD object.” (Sect. 3, “Objects of this class are topologically equivalent to a sphere and comprise most shapes encountered in medical imaging ( e.g. liver, kidneys and lungs). The task is to find a one-to-one mapping which assigns every point on the surface of the mesh a unique position on the unit sphere, described by two parameters longitude θ ∈ [0 .. 2π] and latitude φ ∈ [0 .. π]. The mapping of an arbitrary shape to a sphere inevitably introduces some distortion. There are a number of different approaches which attempt to minimize this distortion, typically preserving either local angles or facet areas while trying to minimize distortions in the other…Due to our optimization strategy (Sect. 4), our focus lies on preserving angles: Moving neighboring points on the parameterization sphere in a specific direction, we expect the corresponding landmarks on the training shape to move in a coherent direction as well” [e.g., a geometric style]. Figure 2 displays the gradient descent optimization method to visualize a minimum distance length influence on an 3D object, e.g., computing for a geometric style gradient and further generates the gradient graphically to create a visual representation. Section 4.2, “this derivation yields a 3D gradient for every landmark, revealing the influence of its movements on the cost function” [for each position included in a first plurality of positions associated with the first 3D CAD object, wherein the first plurality of positions corresponds to a first plurality of sample points for the first 3D CAD object]. Section 4.2, “the scalars uim and vjm are elements of the matrices U and V from (2). Since our MDL cost function uses λm = d2m, we can derive the MDL gradients as”, see equation 9. The partial derivative of equation 9 computes the gradients for every landmark on the 3D object, e.g., computing a different partial derivative of a style comparison metric for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient. On a further note, partial derivative is used to calculate the gradient, and the examiner has interpreted landmarks as positions on the object. Further see Sect. 3 and 4. The examiner has interpreted that generating a derivation that yields a 3D MDL gradient for every landmark using different partial derivatives to find the point on a mesh which is assigned to a position on a sphere and whose landmarks are moved in a specific direction corresponding to the landmarks on the sphere as computing a different partial derivative of a style comparison metric for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient, wherein the first plurality of positions corresponds to a first plurality of sample points for the first 3D CAD object.)
Heimann also teaches “wherein the first geometric style gradient includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point that indicates a direction magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object at the first position”. (Figure 2 displays the gradient descent optimization method to visualize a minimum distance length influence for 3D objects, e.g., geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object. Figure 2 also shows a graphical element, e.g., a vector, represented by an arrow, and a vector is based on both the direction and magnitude of the geometric gradient. Section 4.2, “this derivation yields a 3D gradient for every landmark, revealing the influence of its movements on the cost function” [wherein the first geometric style gradient includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point that indicates a direction and magnitude of a geometric style and at a first position]. Further see Sect. 4. The examiner has interpreted that generating a 3D MDL gradient for every landmark using different partial derivatives and visualizing the gradient as a vector having both a direction and magnitude as wherein the first geometric style gradient includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point that indicates a direction and magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object at the first position.)
Heimann also teaches “generating a first graphical element based on the first vector included in the first geometric style gradient, wherein the first graphical element illustrates a direction and magnitude in which to move a first sample point corresponding to the first position for modifying the first 3D CAD object to increase a geometric style similarity between the first 3D CAD object and the second 3D CAD object at the first position” (Section 4.2, “the scalars uim and vjm are elements of the matrices U and V from (2). Since our MDL cost function uses λm = d2m, we can derive the MDL gradients as”, see equation 9. Section 4.2, “the direction (Δθ, Δφ) for the movement which minimizes the cost function” [a direction included in the first geometric style gradient]. Figure 2 displays the gradient descent optimization method to visualize a minimum distance length influence for the two 3D objects. Figure 2 also shows a graphical element, e.g., a vector, represented by an arrow, and a vector is based on both the direction and magnitude of the geometric gradient. Further see Sect 4. The examiner has interpreted that generating a vector showing the magnitude and direction of the MDL cost function and visualizing that vector on the object as generating a first graphical element based on the first vector included in the first geometric style gradient, wherein the first graphical element illustrates a direction and magnitude in which to move a first sample point corresponding to the first position for modifying the first 3D CAD object to increase a geometric style similarity between the first 3D CAD object and the second 3D CAD object at the first position.)
Heimann also teaches “positioning the first graphical element relative to the first position on the first 3D CAD object within a graphical user interface to generate a first visualization of the first geometric style gradient.” (Section 4.2, “derivation yields a 3D gradient for every landmark” [positions], “revealing the influence of its movements on the cost function.” The examiner has interpreted Figure 2 as showing graphical element (i.e., a vector, represented by an arrow) at each landmark (i.e., position). Heimann also teaches a vector representing the geometric gradient is generated for a plurality of positions of the 3D objected. Further see Sect. 4. The examiner has interpreted generating an output for the vector is positioned on the object extruding outward on every landmark resembling the 3D gradient as positioning the first graphical element relative to the first position on the first 3D CAD object within a graphical user interface to generate a first visualization of the first geometric style gradient.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add the generation of a “[computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric] for each position included in a first plurality of positions [associated with the first 3D CAD object to generate a first geometric style gradient] wherein the first plurality of positions corresponds to a first plurality of sample points for the first 3D CAD object”, “[wherein the first geometric style gradient] includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point [that indicates a direction magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object] at the first position”, “generating a first graphical element based on the first vector included in the first geometric style gradient, wherein the first graphical element illustrates a direction and magnitude in which to move a first sample point corresponding to the first position for modifying the first 3D CAD object to increase a geometric style similarity between the first 3D CAD object and the second 3D CAD object at the first position”, and “positioning the first graphical element relative to the first position on the first 3D CAD object within a graphical user interface to generate a first visualization of the first geometric style gradient,” as conceptually seen from the teaching of Heimann, into that of Ljung because this modification (1) creates a gradient visualization feature at each position for the advantageous purpose of providing insights into the behavior and appearance of the 3D objects as well as creating more details to more precise style definitions for a more accurate comparison for the entire object, (2) uses partial derivatives for gradient creation for the advantageous purpose of optimizing the 3D model when creating the gradient for each position, (3) creates graphical elements for the advantageous purpose of visually displaying the direction or magnitude of the gradient, and (4) positions the graphical elements for the advantageous purpose of visually displaying where the similar or dissimilar gradients take place on the 3D object surface. Further motivation to combine be that Ljung Larhed and Heimann are analogous art to the current claim and directed to creating geometric gradients.
As per Claim 9, Ljung Larhed teaches “computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of the style comparison metric [for each position included in a second plurality of positions] associated with the second 3D CAD object to generate a second geometric style gradient.” (Para. 45, “the search at 906 in one example includes identifying meshes having similar features defined by similar outputs, such as meshes having output values for the relevant properties” [style comparison metrics associated with the second 3D CAD object] “within a predetermined variance of the values for the relevant features of the unknown mesh 902 input into the neural network 900. The trained neural network 900 can be used on the unknown mesh 902 to produce the set of features 904 used to search a database of pre-existing 3D object feature values to identify any similar meshes.” Para. 21, “With the present disclosure, machine learning is used to find an n-dimensional identifier of a 3D object that captures particular features relevant to distinguish that object from other 3D objects. Objects that are similar generate similar feature values” [geometric style and based on the first plurality of style signals and the second plurality of style signals]. Para. 44, “The neural network 812 is adjusted so that the 0.43 and 0.48 output values 810 are brought closer together (i.e., a smaller difference) by causing an increase in the 0.43 value (output) and a decrease in the 0.48 value (output) for this particular feature. For example, the adjusted values may be passed in the reverse direction (backward propagating) to the neural network 812 to converge outputs to the same or similar value” [values brought closer to similar value to show similarity, i.e., geometric style gradient]. Para. 28, “The neural network training system 200 in one example uses back propagation or other training techniques. The neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object ( e.g., 3D mesh) that captures particular features relevant to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects (that are not identical)” [using back propagation on the particular features to generate a n-dimensional identifier and different partial derivatives are the core of back propagation, e.g., computing a different partial derivative associated with the second 3D object to generate a second geometric style gradient]. On a further note, the examiner has interpreted the relevant properties [i.e., features] as styles and the output values of relevant features as the style comparison metrics. Further see Para 21, 28, 44-45, and 59. The examiner has interpreted that distinguishing an object from other 3D objects using particular features through the use of back propagation to find an n-dimensional identifier that is adjusted to increase the output value for the relevant feature properties to converge to a similar value for finding similar objects as computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric associated with the second 3D CAD object to generate a second geometric style gradient.)
Ljung Larhed also teaches “displaying one or more aspects of the second geometric style gradient via the graphical user interface.” (Para. 44, “The neural network 812 is adjusted so that the 0.43 and 0.48 output values 810 are brought closer together (i.e., a smaller difference) by causing an increase in the 0.43 value (output) and a decrease in the 0.48 value (output) for this particular feature. For example, the adjusted values may be passed in the reverse direction (backward propagating) to the neural network 812 to converge outputs to the same or similar value” [one or more aspects of the second geometric style gradient]. Para. 33, “for example, an operator is able to specify the neural network topology using a graphical user interface” [via the graphical user interface]. Para. 63 further teaches “computing apparatus 1102 may comprise an input/output controller” [display] “1118 configured to output information to one or more input devices 1120 and output devices 1122, for example a display or a speaker” [displaying one or more aspects of the second geometric style gradient via the graphical user interface]. Further see Para. 33 and 44. The examiner has interpreted that graphical user interface (GUI) used for displaying outputs of adjusted values through back propagating the neural network to converge outputs to the same or similar value as displaying one or more aspects of the second geometric style gradient via the graphical user interface.)
Ljung Larhed does not specifically teach “[computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of the style comparison metric] for each position included in a second plurality of positions associated with the second 3D CAD object [to generate a second geometric style gradient]”.
However, Heimann teaches a “computing, [based on the first plurality of style signals and the second plurality of style signals,] a different partial derivative of the style comparison metric for each position included in a second plurality of positions associated with the second 3D CAD object to generate a second geometric style gradient.” (Figure 2 displays the gradient descent optimization method to visualize a minimum distance length influence on an 3D object, e.g., computing for a geometric style gradient and further generates the gradient graphically to create a visual representation. Section 4.2, “this derivation yields a 3D gradient for every landmark, revealing the influence of its movements on the cost function” [for each position included in a first plurality of positions associated with the second 3D CAD object]. Section 4.2, “the scalars uim and vjm are elements of the matrices U and V from (2). Since our MDL cost function uses λm = d2m, we can derive the MDL gradients as”, see equation 9. The partial derivative of equation 9 computes the gradients for every landmark on the second 3D object, e.g., computing a different partial derivative of a style comparison metric for each position included in a second plurality of positions associated with the second 3D CAD object to generate a second geometric style gradient. On a further note, partial derivative is used to calculate the gradient, and the examiner has interpreted landmarks as positions on the object. Further see Sect. 4. The examiner has interpreted generating a derivation that yields a 3D MDL gradient for every landmark using different partial derivatives as computing a different partial derivative of a style comparison metric for each position included in a second plurality of positions associated with the second 3D CAD object to generate a second geometric style gradient.)
Furthermore, additionally, Heimann also teaches “displaying one or more aspects of the second geometric style gradient via the graphical user interface.” (Fig. 2, the display of gradients that show both magnitude and direction, graphically. Further see Sect 4. The examiner has interpreted the magnitude and the direction are mere aspects of the gradient.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add the “[computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric] for each position included in a second plurality of positions [associated with the second 3D CAD object to generate a second geometric style gradient]”, as conceptually seen from the teaching of Heimann, into that of Ljung because this modification uses partial derivatives for gradient creation for the advantageous purpose of optimizing the 3D model when creating the gradient for each position. Further motivation to combine be that Ljung Larhed and Heimann are analogous art to the current claim and directed to creating geometric gradients.
As per Claim 10, Ljung Larhed teaches “wherein [the first graphical element comprises an arrow or a line that is centered at] an absolute 3D position corresponding to the first sample point.” (Para. 38, “the orientation and size can be normalized to any rotation or size with the object 500 generally positioned in a center of an evaluation area 508, thereby also normalizing position.” Further see Para. 38. The examiner has interpreted the 3D object being positioned in a center of an area as an absolute 3D position corresponding to the first sample point.)
Ljung Larhed does not does not specifically teach that “first graphical element comprises an arrow or a line”.
However, Heimann teaches “first graphical element comprises an arrow or a line” (Section 4.2, “the direction (Δθ, Δφ) for the movement which minimizes the cost function.” Figure 2 displays the gradient descent optimization method to visualize a minimum distance length influence on an 3D object. Further see Sect. 4. The examiner has interpreted Figure 2 as showing a graphical element, a vector, represented by an arrow as first graphical element comprises an arrow.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add the generation of a “first graphical element comprises an arrow or a line,” as conceptually seen from the teaching of Heimann, into that of Ljung because this modification creates graphical elements as an arrow for the advantageous purpose of visually displaying the direction and magnitude of the gradient. Further motivation to combine be that Ljung Larhed and Heimann are analogous art to the current claim and directed to creating geometric gradients.
Re Claim 11, it is an articles of manufacture claim, having similar limitations of claim 1. Thus, claim 11 is also rejected under the similar rationale as cited in the rejection of claim 1.
Furthermore, regarding claim 11, Ljung Larhed teaches “One or more non-transitory computer readable media including instructions that, when executed by one or more processors” (Para. 62, “computer executable instructions may be provided using any computer-readable media that are accessible by the computing apparatus 1102. Computer-readable media may include, for example, computer storage media such as a memory 1114 and communications media. Computer storage media, such as the memory 1114, include volatile and non-volatile, removable and non-removable media.” Para. 61, “the computing apparatus 1102 comprises one or more processors 1104 which may be microprocessors, controllers or any other suitable type of processors for processing computer executable instructions to control the operation of the electronic device.” Further see Para. 61-62. The examiner has interpreted the processors having computer executable instructions provided by the non-volatile computer-readable media as non-transitory computer readable media including instructions executed by one or more processors.)
As per Claim 15, Ljung Larhed teaches “wherein computing the different partial derivatives of the style comparison metric comprises performing one or more backpropagation operations via the trained neural network.” (Para. 31, “training process including comparing the computed features” [style comparison metrics] “at 216, such as during an iterative back propagation process, wherein the neural network 204 is trained to output similar features for similar meshes” [performing one or more backpropagation operations via the trained neural network]. Further see Para. 31. The examiner has interpreted computing features, the output of the extracted relevant features in iterative back propagation process with a trained neural network as wherein computing the different partial derivatives of the style comparison metric comprises performing one or more backpropagation operations via the trained neural network.)
As per Claim 19, Ljung Larhed teaches “wherein a representation of the first 3D CAD object comprises a boundary representation (B-rep), a 3D point cloud, or a 3D mesh.” (Para. 28, “machine learning to find an n-dimensional identifier of a 3D object (e.g., 3D mesh).” Further see Para. 28. The examiner has interpreted that 3D mesh as a 3D object.)
Re Claim 20, it is a system claim, having similar limitations of claim 1. Thus claim 20 is also rejected under the similar rationale as cited in the rejection of claim 1.
Furthermore, regarding claim 20, Ljung Larhed teaches “A system comprising: one or more memories storing instructions; and one or more processors coupled to the one or more memories that, when executing the instructions”. (Para. 23, “The image processing system 100 includes one or more computers 102 and storage 104 to store meshes (e.g., polygon meshes) and images/videos in some examples” [system]. Para. 62, “computer executable instructions may be provided using any computer-readable media that are accessible by the computing apparatus 1102. Computer-readable media may include, for example, computer storage media such as a memory 1114 and communications media. Computer storage media, such as the memory 1114, include volatile and non-volatile, removable and non-removable media." Further details are also shown in Figure 1. Para. 61, "the computing apparatus 1102 comprises one or more processors 1104 which may be microprocessors, controllers or any other suitable type of processors for processing computer executable instructions to control the operation of the electronic device.” Further see Para. 23 and 61-62. The examiner has interpreted that a computer system having computer executable instructions using computer-readable media that are accessible by processors as a system comprising: one or more memories storing instructions; and one or more processors coupled to the one or more memories that, when executing the instructions.)
As per Claim 21, Ljung Larhed teaches “wherein the first geometric style gradient indicates dissimilarities in shape between the first 3D CAD object and the second 3D CAD object.” (Para. 28, “The neural network training system 200 in one example uses back propagation or other training techniques. The neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object ( e.g., 3D mesh) that captures particular features relevant to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects (that are not identical)” [the first geometric style gradient indicates dissimilarities between the first 3D CAD object and the second 3D CAD object]. Para. 39, “As should be appreciated, the values 600 relate to the particular features of the mesh of the object 500 that are compared when generating a similar mesh (e.g., the object 500 with certain features removed, such as the arm features 606), to train the neural network 204 to perform a fuzzy identification of objects that have similar meshes. It should be appreciated that different values for the 3D object (that has been normalized) are input to the neural network 602 in some examples. These values include, but are not limited to, values relating to volumetric information, shape information” and Para. 87, “The system described above, wherein the computed values correspond to values relating to at least one of volumetric information, shape information, or topology information” [dissimilarities in shape]. Further see Para. 28, 39, 44, 52, 87, 94, 104. The examiner has interpreted that distinguishing an object from other 3D objects using particular features that is adjusted to increase the output value for the relevant feature properties to diverge away from a similar value in the reverse direction for finding objects that are dissimilar where the features correspond to shape information as wherein the first geometric style gradient indicates dissimilarities in shape between the first 3D CAD object and the second 3D CAD object.)
As per Claim 22, Ljung Larhed does not specifically teach “wherein the first geometric style gradient comprises a first plurality of vectors that indicates dissimilarities between geometric styles of the first 3D CAD object and the second 3D CAD object at the first plurality of positions”.
Heimann also teaches “wherein the first geometric style gradient comprises a first plurality of vectors that indicates dissimilarities between geometric styles of the first 3D CAD object and the second 3D CAD object at the first plurality of positions”. (Figure 2 displays the gradient descent optimization method to visualize a minimum distance length influence for 3D objects, e.g., geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object. Figure 2 also shows a graphical element, e.g., a vector, represented by an arrow, and a vector is based on both the direction and magnitude of the geometric gradient. Section 4.2, “this derivation yields a 3D gradient for every landmark, revealing the influence of its movements on the cost function” [wherein the first geometric style gradient comprises a first plurality of vectors that indicates dissimilarities between geometric styles of the first 3D CAD object and the second 3D CAD object at the first plurality of position]. Further see Sect. 4. The examiner has interpreted that generating a 3D MDL gradient for every landmark using different partial derivatives and visualizing the gradient as a vector having both a direction and magnitude as wherein the first geometric style gradient comprises a first plurality of vectors that indicates dissimilarities between geometric styles of the first 3D CAD object and the second 3D CAD object at the first plurality of position.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add the generation of a “wherein the first geometric style gradient comprises a first plurality of vectors that indicates dissimilarities between geometric styles of the first 3D CAD object and the second 3D CAD object at the first plurality of positions,” as conceptually seen from the teaching of Heimann, into that of Ljung because this modification of creating graphical elements for the advantageous purpose of visually displaying the direction or magnitude of the gradient and visually displaying where the similar or dissimilar gradients take place on the 3D object surface. Further motivation to combine be that Ljung Larhed and Heimann are analogous art to the current claim and directed to creating geometric gradients.
Claims 2-3, 5, 7, 12, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed and Heimann as applied to claims 1 and 11 above, and further in view of Bronstein et al. “Geometric Deep Learning: Going beyond Euclidean data” IEEE Signal Processing Magazine Volume: 34, Issue: 4 (July 2017), [herein “Bronstein”].
As per Claim 2, Ljung Larhed teaches “further comprising executing the trained neural network at least once [to map the first 3D CAD object to a first feature map set].” (Para. 5, “the computerized method further comprises training the neural network using the input mesh and the plurality of training meshes by tuning output of the neural network to identify similar non-identical meshes” [executing the trained neural network at least once]. Further see Para. 5. The examiner has interpreted that identifying similar non-identical meshes by training neural network as executing the trained neural network at least once.)
Ljung Larhed nor Heimann teach “[executing the trained neural network at least once] to map the first 3D CAD object to a first feature map set.”
However, in the same field of endeavor namely using neural networks to find geometric characteristics of 3D objects, Bronstein teaches “executing the trained neural network at least once to map the first 3D CAD object to a first feature map set.” (Sect. 3 Para. 7, “CNN” [neural network] “consists of several convolutional layers of the form g = Cr(f), acting on a p-dimensional input f(x) = (f1 (x), ... , fp(x)) by applying a bank of filters Γ = (ϒl,l'), l = 1, ..., q, l' = 1, ..., p and point-wise non-linearity ξ [equation (6)] producing a q-dimensional output g(x) = (g1(x),..., gq(x)) often referred to as the feature maps” [3D objects are mapped as a feature map using a convolutional neural network, e.g., executing the trained neural network at least once to map the first 3D CAD object to a first feature map set]. Fig. 1 shows that these functions and equations are applied to 3D CAD objects. Further see Sect. 3-4. The examiner is interpreting producing the feature maps for an output for 3D objects through the use of a convolutional neural network as executing the trained neural network at least once to map the first 3D CAD object to a first feature map set.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “executing the trained neural network at least once to map the first 3D CAD object to a first feature map set” as conceptually seen from the teaching of Bronstein, into that of Ljung Larhed and Heimann combined because this modification maps the 3D object into the map set for the advantageous purpose of transforming a 3D object into a model for the system to mathematically compare to other 3D objects or meshes. Further motivation to combine be that Ljung Larhed, Heimann, and Bronstein are analogous art to the current claim and directed to creating geometric gradients.
As per Claim 3, Ljung Larhed teaches “wherein [a first feature map included in the first feature map set has] a plurality of activations associated with a same layer of the trained neural network.” (Para. 2, “a neural network is a collection of layers of nodes interconnected by edges and where weights which are learned during a training phase are associated with the nodes. Input features are applied to one or more input nodes of the network and propagate through the network in a manner influenced by the weights (the output of a node is related to the weighted sum of the inputs). As a result, activations at one or more output nodes of the network are obtained.” Further see Para. 2. The examiner has interpreted that the layers of the neural network contain nodes where each node contains a plurality of activations as wherein a plurality of activations associated with a same layer of the trained neural network.)
Ljung Larhed nor Heimann specifically teach a “wherein a first feature map included in the first feature map set [has a plurality of activations associated with a same layer of the trained neural network].”
However, Bronstein teaches “wherein a first feature map included in the first feature map set has a first feature map included in the first feature map set”. (Page 4 Para. 7, “CNN” [neural network] “consists of several convolutional layers of the form g = Cr(f), acting on a p-dimensional input f(x) = (f1 (x), ... , fp(x)) by applying a bank of filters Γ = (ϒl,l'), l = 1, ..., q, l' = 1, ..., p and point-wise non-linearity ξ [equation (6)] producing a q-dimensional output g(x) = (g1(x),..., gq(x)) often referred to as the feature maps” [3D objects are mapped as a feature map]. Fig. 1 shows that these functions and equations are applied to 3D CAD objects. Further see Sect. 3-4. The examiner is interpreting producing the feature maps for an output for 3D objects through the use of convolutional layers of a convolutional neural network as wherein a first feature map included in the first feature map set has a plurality of activations associated with a same layer of the trained neural network.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “wherein a first feature map included in the first feature map set [has a plurality of activations associated with a same layer of the trained neural network]” as conceptually seen from the teaching of Bronstein, into that of Ljung Larhed and Heimann combined because this modification maps the 3D object into the map set for the advantageous purpose of transforming a 3D object into a model for the system to mathematically compare to other 3D models or meshes. Further motivation to combine be that Ljung Larhed, Heimann, and Bronstein are analogous art to the current claim and directed to creating geometric gradients.
Re Claim 5, it is a process claim, having similar limitations of claim 2. Thus claim 5 is also rejected under the similar rationale as cited in the rejection of claim 2.
As per Claim 7, Ljung Larhed teaches “wherein computing the different partial derivatives of the style comparison metric [is further based on the first feature map set and the second feature map set].” (Para. 45, “the search at 906 in one example includes identifying meshes having similar features defined by similar outputs, such as meshes having output values for the relevant properties within a predetermined variance of the values for the relevant features” [style comparison metrics] “of the unknown mesh 902 input into the neural network 900. The trained neural network 900 can be used on the unknown mesh 902 to produce the set of features 904 used to search a database of pre-existing 3D object feature values to identify any similar meshes.” Para. 44, “The neural network 812 is adjusted so that the 0.43 and 0.48 output values 810 are brought closer together (i.e., a smaller difference) by causing an increase in the 0.43 value (output) and a decrease in the 0.48 value (output) for this particular feature. For example, the adjusted values may be passed in the reverse direction (backward propagating) to the neural network 812 to converge outputs to the same or similar value” [values brought closer to similar value to show similarity, i.e., geometric style gradient of the style comparison metric]. Para. 28, “The neural network training system 200 in one example uses back propagation or other training techniques. The neural network training system 200 includes a training processor 202 that uses machine learning to find an n-dimensional identifier of a 3D object ( e.g., 3D mesh) that captures particular features relevant to distinguish the 3D object from other 3D objects, such that a neural network 204 is trained to generate similar feature values for similar objects (that are not identical)” [using back propagation on the particular features to generate a n-dimensional identifier and different partial derivatives are the core of back propagation, e.g., computing a different partial derivatives of the style comparison metric. On a further note, the examiner has interpreted the relevant properties [i.e., features] as styles and the output values of relevant features as the style comparison metrics. Further see Para 21, 28, 44-45, and 59. The examiner has interpreted that using of back propagation to find an n-dimensional identifier that is adjusted to increase the output value for the relevant feature properties to converge to a similar value for finding similar objects as computing the different partial derivatives of the style comparison metric.)
Furthermore, Heimann also teaches “computing the different partial derivatives of the style comparison metric”. (Section 4.2, “the scalars uim and vjm are elements of the matrices U and V from (2). Since our MDL cost function uses λm = d2m, we can derive the MDL gradients as”, see equation 9. The partial derivative of equation 9 computes the gradients for every landmark on the 3D object, e.g., computing a different partial derivative of a style comparison metric. Further see Sect. 4. The examiner has interpreted generating a derivation that yields a 3D MDL gradient for every landmark using different partial derivatives as computing the different partial derivatives of the style comparison metric.)
Neither Ljung Larhed nor Heimann specifically teach “[wherein computing the different partial derivatives of the style comparison metric] further based on the first feature map set and the second feature map set.”
However, Bronstein teaches “[wherein computing the different partial derivatives of the style comparison metric] further based on the first feature map set and the second feature map set.” (Page 4 Para. 7, “CNN” [neural network] “consists of several convolutional layers of the form g = Cr(f), acting on a p-dimensional input f(x) = (f1 (x), ... , fp(x)) by applying a bank of filters Γ = (ϒl,l'), l = 1, ..., q, l' = 1, ..., p and point-wise non-linearity ξ [equation (6)] producing a q-dimensional output g(x) = (g1(x),..., gq(x)) often referred to as the feature maps” [3D objects are mapped as a feature map]. Fig. 1 shows that these functions and equations are applied to 3D CAD objects. Further see Sect. 3-4. The examiner is interpreting producing the feature maps for an output for 3D objects through the use of convolutional layers of a convolutional neural network as further based on the first feature map set and the second feature map set.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “[computing the different partial derivatives of the style comparison metric] is further based on a first feature map included in the first feature map set” as conceptually seen from the teaching of Bronstein, into that of Ljung Larhed and Heimann combined because this modification maps the 3D object into the map set for the advantageous purpose of transforming a 3D object into a model for the system to mathematically compare to other 3D models or meshes as well as optimizing the performance of the system by not adding a significant computational overload to the system. Further motivation to combine be that Ljung Larhed, Heimann, and Bronstein are analogous art to the current claim and directed to creating geometric gradients.
Re Claim 12, it is an articles of manufacture claim, having similar limitations of claim 2. Thus claim 12 is also rejected under the similar rationale as cited in the rejection of claim 2.
Re Claim 14, it is an articles of manufacture claim, having similar limitations of claim 5. Thus claim 14 is also rejected under the similar rationale as cited in the rejection of claim 5.
Claims 4, 6, and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed, Heimann, and Bronstein as applied to claims 2 and 12 above, and further in view of US Patent 9,922,432 B1 Risser [herein “Risser”].
Per Claim 4, Ljung Larhed teaches “generating the first plurality of style signals [comprises extracting second-order activation information from one or more feature maps included in the first feature map set].” (Para. 43, “neural network 812 is adjusted so that the output values 810 for the relevant features” [style signals] “converge to the same value or a value within a defined threshold variance that allows for subsequent identification of the first and second meshes 800....neural network 812 is adjusted so that the output values 810 for the relevant features of the first mesh 800 and the third mesh 806 {which are not perceptually similar} diverge to be further apart.” Ljung Larhed teaches a method for creating relevant features values of 3D objects, where similar values to converge and dissimilar values to diverge. The examiner has interpreted style to be any relevant features of the 3D object, such as scale, position and orientation, see Para. 21.)
Ljung Larhed nor Heimann specifically teach “[generating the first plurality of style signals comprises extracting second-order activation information from] one or more feature maps included in the first feature map set.”
However, Bronstein teaches “one or more feature maps included in the first feature map set.” (Page 4 Para. 7, “CNN” [neural network] “consists of several convolutional layers of the form g = Cr(f), acting on a p-dimensional input f(x) = (f1 (x), ... , fp(x)) by applying a bank of filters Γ = (ϒl,l'), l = 1, ..., q, l' = 1, ..., p and point-wise non-linearity ξ [equation (6)] producing a q-dimensional output g(x) = (g1(x),..., gq(x)) often referred to as the feature maps” [3D objects are mapped as a feature map]. Fig. 1 shows that these functions and equations are applied to 3D CAD objects. Further see Sect. 3-4. The examiner is interpreting producing the feature maps for an output for 3D objects through the use of convolutional layers of a convolutional neural network as one or more feature maps included in the first feature map set.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “[generating the first plurality of style signals comprises extracting second-order activation information from] one or more feature maps included in the first feature map set” as conceptually seen from the teaching of Bronstein, into that of Ljung Larhed and Heimann combined because this modification maps the 3D object into the map set for the advantageous purpose of transforming a 3D object into a model for the system to mathematically compare to other 3D objects or meshes. Further motivation to combine be that Ljung Larhed, Heimann, and Bronstein are analogous art to the current claim and directed to creating geometric gradients.
Neither Ljung Larhed, Heimann nor Bronstein specifically teach “[generating the first plurality of style signals] comprises extracting second-order activation information [from one or more feature maps included in the first feature map set].”
However, in the same field of endeavor namely using neural networks to transfer style, Risser teaches “comprises extracting second-order activation information.” (Col. 12 lines 31-37, “by subtracting off the mean activation before computing inner products, covariance matrices explicitly preserve statistical moments of various orders in the parametric model. By this we explicitly refer to the mean of all feature vectors as the first order moment and to the co-activations of feature vectors centered around their mean as second order moments.” Further see Col. 12. The examiner interprets the first and second order moments of the vector co-activations as the first and second order activation information.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “comprises extracting second-order activation information” as conceptually seen from the teaching of Risser, into that of Ljung Larhed, Heimann, and Bronstein combined because this modification of using the second order for the advantageous purpose to preserve statistical moments of various orders in the parametric model. Further motivation to combine be that Ljung Larhed, Heimann, Bronstein, and Risser are analogous art to the current claim and directed to using neural networks to transfer style.
Re Claim 6, it is a process claim, having similar limitations of claim 4. Thus claim 6 is also rejected under the similar rationale as cited in the rejection of claim 4.
Re Claim 13, it is an articles of manufacture claim, having similar limitations of claim 4. Thus claim 13 is also rejected under the similar rationale as cited in the rejection of claim 4.
Claims 8 and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed and Heimann as applied to claim 1, and further in view of US 2007/0055401 A1 Van Bael et al. [herein “Van Bael”].
As per Claim 8, Ljung Larhed teaches “wherein each position included in the first plurality of positions comprises an absolute 3D position in a geometric domain [and corresponds to a different sample point in a UV domain].” (Para. 38, “the orientation and size can be normalized to any rotation or size with the object 500 generally positioned in a center of an evaluation area 508, thereby also normalizing position.” Further see Para. 38. The examiner has interpreted the 3D object being position in a center of an area as an absolute 3D position.)
Ljung Larhed nor Heimann teach “corresponds to a different sample point in a UV domain.”
However, in the same field of endeavor namely computing 3D Objects graphically, Van Bael teaches “corresponds to a different sample point in a UV domain.” (Para. 105, “FIG. 22 shows the two points 2103 and 2105 and the line 2107 on the 2D unfolded view of the carton after the (u,v) coordinates have been determined.” Further see Para. 105. The examiner has interpreted that Figure 21 and Figure 22 show that points in the UV domain correspond to the 3D domain.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add that the absolute 3D position “corresponds to a different sample point in a UV domain” as conceptually seen from the teaching of Van Bael, into that of Ljung Larhed and Heimann combined because this modification of translating the 3D into the 2D for the advantageous purpose to easily visualize aspects and features of 3D objects in the 2D domain to easily compare and visualize the distance variations of the surface of two 3D objects for detecting style changes or similarities. Further motivation to combine be that Ljung Larhed, Heimann, and Van Bael are analogous art to the current claim and directed to computing 3D Objects graphically.
As per Claim 24, neither Ljung Larhed nor Heimann specifically teach “wherein the first sample point comprises a first UV sample point”.
However, Van Bael teaches “wherein the first sample point comprises a first UV sample point”. (Para. 105, “FIG. 22 shows the two points 2103 and 2105 and the line 2107 on the 2D unfolded view of the carton after the (u,v) coordinates have been determined.” Further see Para. 105. The examiner has interpreted that Figure 21 and Figure 22 show that points in the 3D domain correspond to the UV domain correspond.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add that the absolute 3D position “wherein the first sample point comprises a first UV sample point” as conceptually seen from the teaching of Van Bael, into that of Ljung Larhed and Heimann combined because this modification of translating the 3D into the 2D for the advantageous purpose to easily visualize aspects and features of 3D objects in the 2D domain to easily compare and visualize the distance variations of the surface of two 3D objects for detecting style changes or similarities. Further motivation to combine be that Ljung Larhed, Heimann, and Van Bael are analogous art to the current claim and directed to computing 3D Objects graphically.
Claims 18 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Ljung Larhed and Heimann as applied to claim 11 above, and further in view of US Patent 9,922,432 B1 Risser [herein “Risser”].
As per Claim 18, Ljung Larhed teaches “wherein the style comparison metric [comprises a similarity metric or a distance metric].” (Para. 45, "the search at 906 in one example includes identifying meshes having similar features defined by similar outputs, such as meshes having output values for the relevant properties” [style comparison metrics] “within a predetermined variance of the values for the relevant features of the unknown mesh 902 input into the neural network 900. The trained neural network 900 can be used on the unknown mesh 902 to produce the set of features 904 used to search a database of pre-existing 3D object feature values to identify any similar meshes.” Further see Para. 45. The examiner has interpreted the relevant properties [i.e., features] as styles and the output values of relevant features as the style comparison metrics.)
Ljung Larhed nor Heimann does not does not specifically teach “wherein style comparison metric comprises a similarity metric or a distance metric.”
However, in the same field of endeavor namely using neural networks to transfer style, Risser teaches that the “wherein style comparison metric comprises a similarity metric or a distance metric.” (Col. 10, line 12-15, “content loss can be included as well, where the content loss is some distance metric between raw neural activations calculated for the content image and the image being synthesized.” Further see Col. 10 The examiner has interpreted the content loss is a style comparison metric that is based on the distance.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “wherein style comparison metric comprises a similarity metric or a distance metric” as conceptually seen from the teaching of Risser, into that of Ljung Larhed and Heimann combined because this modification of using a distance metric for the advantageous purpose to compare the distance variations of the surface of two 3D objects in detecting style changes or similarities. Further motivation to combine be that Ljung Larhed, Heimann, and Risser are analogous art to the current claim and directed to using neural networks to transfer style.
As per Claim 23, neither Ljung Larhed nor Heimann specifically teach “inputting, to the trained neural network, a first UV-grid comprising a grid of samples associated with the first 3D CAD object in the UV domain; and generating, via execution of the trained neural network, the first plurality of style signals further based on the first UV-grid”.
However, Risser teaches “inputting, to the trained neural network, a first UV-grid comprising a grid of samples associated with the first 3D CAD object in the UV domain; and generating, via execution of the trained neural network, the first plurality of style signals further based on the first UV-grid”. (Col. 37 Ln. 26-39, “3D models typically contain UV coordinates at each vertex which define the 2D parameterization of the 3D surface” and Col. 38 Ln. 11-21 “Processes in accordance with some of these embodiments introduce an underlying vector field that frames the local orientation around each pixel. As CNNs work by performing convolution across an image, the vector field directs the local orientation of the convolution. Thus, these processes can bi-linearly interpolate sampling of neural activations from the previous layer. Where the convolution kernel extends beyond the scope of an atlas chart, the gutter space of pointers redirects to another atlas chart. During the back-propagation phase of the process, inverse mapping can be used in a manner similar to what is described above with respect to convolution. This allows these processes to perform CNN image synthesis directly in UV space for on-model synthesis” [inputting, to the trained neural network, a first UV-grid comprising a grid of samples associated with the first 3D CAD object in the UV domain]. Col. 15 Ln. 43-44, “A process for providing CNN-based image synthesis that performs style transfer using localized loss functions” [i.e., generating, via execution of the trained neural network, the first plurality of style signals further based on the first UV-grid]. Further see Col. 6, 9-10, 15, and 37-38. The examiner has interpreted that defining a 2D parameterization with UV coordinates of the 3D surface model to perform an image synthesis to evaluate a loss of style transfer by performing convolution on the image that directs the vector field to the local orientation on across the samples directly in the UV space for the model using a Convolutional Neural Network (CNN) as inputting, to the trained neural network, a first UV-grid comprising a grid of samples associated with the first 3D CAD object in the UV domain; and generating, via execution of the trained neural network, the first plurality of style signals further based on the first UV-grid.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “inputting, to the trained neural network, a first UV-grid comprising a grid of samples associated with the first 3D CAD object in the UV domain; and generating, via execution of the trained neural network, the first plurality of style signals further based on the first UV-grid” as conceptually seen from the teaching of Risser, into that of Ljung Larhed and Heimann combined because this modification of using the UV domain to generate style transfer metrics for the advantageous purpose to compare the distance variations of the surface of two 3D objects in detecting style changes or similarities. Further motivation to combine be that Ljung Larhed, Heimann, and Risser are analogous art to the current claim and directed to using neural networks to transfer style.
Response to Arguments
Applicant’s arguments, see Pg. 10-16, filed July 31, 2026, with respect to the rejection(s) of the claims under 35 U.S.C. § 101 have been fully considered and are persuasive with regards to the amended independent claims that integrate the claimed invention into a practical application. Therefore, the rejection has been withdrawn.
Applicant's arguments filed on July 31, 2026 have been fully considered but they are not persuasive with respect to the rejection(s) of the claims under 35 U.S.C. § 103.
Applicant argues that the combination of references does not teach each and every limitation in the amended claim 1 because cited references fail to teach “computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient, wherein the first plurality of positions corresponds to a first plurality of sample points for the first 3D CAD object”, “wherein the first geometric style gradient includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point that indicates a direction and magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object at the first position”, and “generating a first graphical element based on the first vector included in the first geometric style gradient, wherein the first graphical element illustrates a direction and magnitude in which to move a first sample point corresponding to the first position for modifying the first 3D CAD object to increase a geometric style similarity between the first 3D CAD object and the second 3D CAD object at the first position” (See Applicant’s response, Pg. 15-18).
MPEP 2145(IV) recites “one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references.” Applicant’s reply fails to address the combined teaching of the applied references and instead only argues that each reference individually does not teach all of the claim limitations. One cannot show nonobviousness by attacking reference individually where the rejections are based on combinations of references. MPEP § 2143.03 recites “All words in a claim must be considered in judging the patentability of that claim against the prior art” and “Examiners must consider all claim limitations when determining patentability of an invention over the prior art.”
For the limitation of “computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient, wherein the first plurality of positions corresponds to a first plurality of sample points for the first 3D CAD object”, Ljung Larhed discloses computing, based on the first plurality of style signals and the second plurality of style signals, a different partial derivative of a style comparison metric associated with the first 3D CAD object as distinguishing an object from other 3D objects using particular features through the use of back propagation to find an n-dimensional identifier that is adjusted to increase the output value for the relevant feature properties to converge to a similar value for finding similar objects, as provided in the rejection above. It should be noted that adjusting the values using backward propagation so the values make the objects more similar is a different partial derivative of a style comparison metric to generate a first geometric style gradient. Ljung Larhed does not disclose computing these derivatives for each position on the object. Heimann discloses “computing a different partial derivative of a style comparison metric for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient, wherein the first plurality of positions corresponds to first plurality of sample points for the first 3D CAD object” as generating a derivation that yields a 3D MDL gradient for every landmark using different partial derivatives to find the point on a mesh which is assigned to a position on a sphere and whose landmarks are moved in a specific direction corresponding to the landmarks on the sphere, as accomplished through the differential partial equations, Equ. 9, and gradients seen in Fig. 2. Points on the sphere are moved in a specific direction to the corresponding landmarks on the training shape through the use of the derivative, e.g., a geometric style comparison metric and this is done for each landmark on the sphere. For the limitation of “wherein the first geometric style gradient includes a first vector associated with a first position included in the first plurality of positions corresponding to a first sample point that indicates a direction and magnitude of a geometric style dissimilarity between the first 3D CAD object and the second 3D CAD object at the first position”, Heimann discloses generating a 3D MDL gradient for every landmark using different partial derivatives and visualizing the gradient as a vector having both a direction and magnitude as accomplished through the differential partial equations, Equ. 9, and gradients seen in Fig. 2 as vectors. Performing the gradient for every landmark includes the first position.
For the limitation of “generating a first graphical element based on the first vector included in the first geometric style gradient, wherein the first graphical element illustrates a direction and magnitude in which to move a first sample point corresponding to the first position for modifying the first 3D CAD object to increase a geometric style similarity between the first 3D CAD object and the second 3D CAD object at the first position” is also taught by Heimann through the generation of a vector showing the magnitude and direction of the MDL cost function and visualizing that vector for each landmark. The examiner would like to refer to Fig. 2 of Heimann which displays the gradient descent optimization method to visualize a minimum distance length influence for the two 3D objects and also shows a graphical element, e.g., a vector, represented by an arrow, and a vector is based on both the direction and magnitude of the geometric gradient.
By combining the teaching of computing a different partial derivative for each position included in a first plurality of positions associated with the first 3D CAD object to generate a first geometric style gradient represented as a vector showing magnitude and direction, as mapped above, of Heimann into Ljung Larhed the claimed limitation is taught. For example, by inserting the Heimann teaching of computing a different partial derivative for each position of a 3D CAD object to generate a first geometric style gradient which moves the point of the sphere to the specific direction of the training shape as taught by Heimann into the Ljung Larhed teaching of computing the derivative of the style comparison metric to generate the geometric style gradient at each location on the object to reflect both the increase and decrease of style similarity can be generate the claimed limitations. Additional emphasis and citations have been added to this mapping in the rejection above to the amended limitations.
Therefore, all of the limitations of the amended claim 1 are disclosed in either Ljung Larhed and Heimann, and the combination of these references renders the claimed invention obvious. Therefore, applicant’s arguments are not persuasive and the rejection of claim 1 as obvious over Ljung Larhed in view of Heimann is maintained.
As provided in the summary of the interview conducted on August 19, 2026, the examiner recommended incorporating the subject matter of both claims 19 and 23 into the independent claims to overcome the 35 USC § 103 rejection. Specifically, the examiner recommended to amend the independent claims to include the limitations of claim 23 in combination with claim 19 as “wherein generating the first plurality of style signals comprises: inputting, to the trained neural network, a first UV-grid comprising a grid of samples associated with the first 3D CAD object in an UV domain; and generating, via execution of the trained neural network, the first plurality of style signals further based on a boundary representation (B-rep) of the first 3D CAD object in the first UV-grid”. Since this would incorporate subject matter not taught by Ljung Larhed or Heimann, this would overcome the 103 rejection. Further, a rejection over Ljung Larhed and Heimann in view of Risser could not be applied since Risser, while teaching a UV domain to evaluate style loss, does not teach the B-rep of the first 3D CAD object in the UV domain for the generation of the style loss.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 10,789,622 B2 Ayush; Kumar et al. teaches a method for generating similar 3D objects by determining the style compatibility of their geometric feature vectors and recommending objects based on the style compatibility .
Examiner’s Note: The examiner has cited particular columns and line numbers in the reference that applied to the claims above for the convenience of the applicant. Although the specified citations are representative of the art and are applied to specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested from the applicant, to fully consider the references in their entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner. In the case of amending the claimed invention, the applicant is respectfully requested to indicate the portion(s) of the specification which dictate(s) the structure relied on for the proper interpretation and also to verify and ascertain the metes and bound of the claimed invention.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Simeon P Drapeau whose telephone number is (571)-272-1173. The examiner can normally be reached Monday - Friday, 8 a.m. - 5 p.m. ET.
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, Ryan Pitaro can be reached on (571) 272-4071. 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.
/SIMEON P DRAPEAU/Examiner, Art Unit 2188
/RYAN F PITARO/Supervisory Patent Examiner, Art Unit 2188
Prosecution Insights