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 .
Response to Arguments
Applicant’s arguments, see p. 11, filed December 23, 2025, with respect to the objection have been fully considered and are persuasive. The objection to Claim 2 has been withdrawn.
Applicant’s arguments with respect to claim(s) 1-31 and 33-56 have been considered but are moot because new grounds of rejection are made in view of Tsai (US012254681B2).
Applicant argues that none of the excerpts in Bermudez (US 20200210636A1) relied on by the Examiner describe specifically performing a test time optimization of a representation of a 3D object from a 3D free-form sketch of the 3D object (p. 12, 1st paragraph).
In reply, the Examiner points out that this limitation was previously in Claim 32 which has now been cancelled. Tsai was relied upon to teach this limitation, not Bermudez.
Applicant argues that Bermudez’s method does not involve optimizing an existing representation of a 3D object or updating the existing representation of the 3D object based on a loss computed between the denoised 2D image and the first 2D image, but instead includes using a computed loss during supervised training to update weights of the neural network (p. 12, 2nd paragraph).
In reply, the Examiner points out that Applicant’s specification describes “the representation of the 3D object is updated based on a loss computed between the denoised 2D image and the first 2D image…updating the representation of the 3D object may include adjusting weights” ([0028-0029], p. 7-8). Applicant’s specification describes “compares resulting outputs against a set of expected or desired outputs…adjust weights to refine an output of untrained neural network 706 using a loss function” ([0067], p. 18-19). Thus, Applicant’s specification teaches updating the representation of the 3D object based on a loss computed between the denoised 2D image and the first image by using a loss function to adjust weights. Sanchez Bermudez teaches the loss may penalize quantities e.g. each via a respective loss term. Penalized quantities may comprise, for a time step respective to each sequence curve and for the step respective to the sweep curve (each freehand drawing being here associated to a respective ground truth solid CAD feature) [0149]. By minimizing such a loss, act on the weights so as to tend to make the respective continuous parameters values outputted by each RNN cell and the part close to their ground truth values. The loss penalizes a disparity of the inferred curve sequence length with its ground truth value [0153]. Define a loss comparing the input drawing to the obtained rendered image [0270]. Machine-learning process may allow to propose plausible solid CAD features, or geometries produced by a swept 2D sketch, the predictions of the machine-learning process are robust to noise [0279]. Thus, Sanchez Bermudez teaches a loss is computed between the denoised 2D image and the first 2D image, and the computed loss is used to adjust weights to make the representation of the 3D object close to the first 2D image so that it updates the representation of the 3D object based on the loss computed between the denoised 2D image and the first 2D image. Thus, this is similar to what is described in Applicant’s specification, as discussed above.
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.
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.
Claim(s) 1-4, 8, 11-13, 16, 18, 19, 23, 29, 33-38, 40, 44, 46-50, 54, and 56 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) in view of Tsai (US012254681B2).
As per Claim 1, Sanchez Bermudez teaches a method, comprising: at a device, performing a optimization of a representation of a three-dimensional (3D) object from a 3D free-form sketch of the 3D object (neural network is configured for inference, from a freehand drawing representing a 3D shape, of a solid CAD feature representing the 3D shape, [0027]) by: rendering a first two-dimensional (2D) image from a specified viewpoint of the representation of the 3D object (providing a viewpoint (a 3D position in the 3D space where the solid CAD feature is defined) from which the respective 3D shape is visible, and rendering in an image edges based on the viewpoints, [0129]); adding noise to the first 2D image to generate a noisy 2D image (each edge may be represented by respective parametric curves, having control points, Bezier curves are represented in a 2D image, adding a noise to parameters of the at least one respective parametric curve, the noise may move the control points, [0134], Gaussian noise is added to each parameter of the curve, [0234]); rendering a second 2D image from the specified viewpoint of the 3D free-form sketch of the 3D object (freehand drawing representing a 3D shape is a 2D image designable freehand by a user for representing the 3D shape, by designable freehand, it is meant that the drawing presents characteristics of a 2D image sketched by a user, [0075]); using a pretrained 2D sketch-to-2D image model to denoise the noisy 2D image with the second 2D image as a control signal, wherein an output of the pretrained 2D sketch-to-2D image model is a denoised 2D image; and updating the representation of the 2D object based on a loss computed between the denoised 2D image and the first 2D image (learnt neural network is usable for converting a freehand drawing into a solid CAD feature, the value of a 2D similarity evaluated between a 2D projection of the result of executing the solid CAD feature and the freehand drawing, [0091], with respect to a candidate solid CAD feature for inference, the loss may be invariant to a cyclic permutation of its curve sequence and invariant to replacing the closed 2D planar sketch by its image at the end of the sweep operation and inverting the sweep operation, such a loss handles the case of a freehand drawing possibly and correctly representing different solid CAD features if the solid CAD features yield a same 3D shape, [0148], the loss may penalize quantities e.g. each via a respective loss term, penalized quantities may comprise, for a time step respective to each sequence curve and for the step respective to the sweep curve (each freehand drawing being here associated to a respective ground truth solid CAD feature), [0149], by minimizing such a loss, act on the weights so as to tend to make the respective continuous parameters values outputted by each RNN cell and the part close to their ground truth values, the loss penalizes a disparity of the inferred curve sequence length with its ground truth value, [0153], define a loss comparing the input drawing to the obtained rendered image, [0270], once the model is learned, the user may establish a new freehand drawing and feed the model with it, the most probable solid CAD feature may then be generated, [0278], machine-learning process may allow to propose plausible solid CAD features, or geometries produced by a swept 2D sketch, the predictions of the machine-learning process are robust to noise, [0279]). Applicant’s specification describes “the representation of the 3D object is updated based on a loss computed between the denoised 2D image and the first 2D image…updating the representation of the 3D object may include adjusting weights” ([0028-0029], p. 7-8). Applicant’s specification describes “compares resulting outputs against a set of expected or desired outputs…adjust weights to refine an output of untrained neural network 706 using a loss function” ([0067], p. 18-19). Thus, Applicant’s specification teaches updating the representation of the 3D object based on a loss computed between the denoised 2D image and the first image by using a loss function to adjust weights. Sanchez Bermudez teaches the loss may penalize quantities e.g. each via a respective loss term. Penalized quantities may comprise, for a time step respective to each sequence curve and for the step respective to the sweep curve (each freehand drawing being here associated to a respective ground truth solid CAD feature) [0149]. By minimizing such a loss, act on the weights so as to tend to make the respective continuous parameters values outputted by each RNN cell and the part close to their ground truth values. The loss penalizes a disparity of the inferred curve sequence length with its ground truth value [0153]. Define a loss comparing the input drawing to the obtained rendered image [0270]. Machine-learning process may allow to propose plausible solid CAD features, or geometries produced by a swept 2D sketch, the predictions of the machine-learning process are robust to noise [0279]. Thus, Sanchez Bermudez teaches a loss is computed between the denoised 2D image and the first 2D image, and the computed loss is used to adjust weights to make the representation of the 3D object close to the first 2D image so that it updates the representation of the 3D object based on the loss computed between the denoised 2D image and the first 2D image. Thus, this is similar to what is described in Applicant’s specification, as discussed above.
However, Sanchez Bermudez does not teach that the optimization is a test time optimization. However, Tsai teaches that the optimization is a test time optimization (Test-Time Adaptation can enable the model to quickly adapt to new target test data without having the source domain data, an entropy minimization method can be utilized to optimize for test-time batch norm parameters without involving a proxy task during training, col. 7, lines 38-42).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez so that the optimization is a test time optimization as suggested by Tsai. It is well-known in the art that this allows models to better align with human preferences or complex task structures without the need for costly retraining.
As per Claim 2, Sanchez Bermudez teaches wherein the representation of the 3D object is one of: a neural radiance field (NeRF) model, a signed distance function, a mesh, or a Gaussian Splatting representation (3D graphical representation may comprise a 3D tessellation of the outer surface of the solid, a mesh tessellation, [0062]).
As per Claim 3, Sanchez Bermudez teaches wherein the 3D free-form sketch of the 3D object is manually generated by a user [0075].
As per Claim 4, Sanchez Bermudez teaches wherein the pretrained 2D sketch-to-2D image model is a machine learning model pretrained on pairs of 2D sketches and 2D images (formed dataset comprises training samples each including a freehand drawing of a respective 3D shape and a respective solid CAD feature representing the 3D shape, the dataset is thus configured for learning a neural network usable for converting a freehand drawing into a solid CAD feature, potential inputs of the neural networks (freehand drawings) are each labeled in the dataset by a ground truth output (a respective solid CAD feature), [0089], providing a dataset including freehand drawings each representing a respective 3D shape, and learning the neural network based on the dataset, [0090], learnt neural network is usable for converting a freehand drawing into a solid CAD feature, the value of a 2D similarity evaluated between a 2D projection of the result of executing the solid CAD feature and the freehand drawing, [0091]).
As per Claim 8, Sanchez Bermudez teaches wherein updating the representation of the 3D object includes: adjusting weights of the representation of the 3D object (by minimizing such a loss, act on the weights of the neural network so as to tend to make the respective continuous parameter values outputted by each RNN cell and the part close to their ground truth values, [0153], the loss we minimize to train the weights of the network, [0257], [0258], minimizing this loss allows use to optimize the weights of the networks, [0272]).
As per Claim 11, Sanchez Bermudez teaches wherein a result of the optimizing is an optimized representation of the 3D object (loss for a supervised training which lead to an accurate result are now discussed, [0147], this improves accuracy of the learning, [0148], [0027]).
As per Claim 12, Sanchez Bermudez teaches wherein the optimized representation of the 3D object is renderable from a user-selected viewpoint for presentation to the user ([0075], freehand drawing is thus a 2D image data structure which comprises 2D planar strokes representing a 3D shape from a respective viewpoint from where the 3D shape is visible, the strokes are defined in a same plane and represent a perspective of the 3D shape from the respective viewpoint, each stroke is a continuous curve defined in the plane, [0076], [0027]).
As per Claim 13, Claim 13 is similar in scope to Claim 1, and therefore is rejected under the same rationale.
As per Claim 16, Sanchez Bermudez teaches wherein the representation of the 3D object is a mesh [0062].
As per Claim 18, Claim 18 is similar in scope to Claim 3, and therefore is rejected under the same rationale.
As per Claim 19, Sanchez Bermudez teaches wherein the defined camera position is a randomly sampled camera position (the viewpoint may be provided in any manner, for example sampled (randomly) in a set of candidate viewpoints, [0130]).
As per Claims 23, 29, 33, and 34, these claims are similar in scope to Claims 4, 8, 11, and 12 respectively, and therefore are rejected under the same rationale.
As per Claim 35, Claim 35 is similar in scope to Claim 13, except the Claim 35 is directed to a system, comprising: a non-transitory memory storage comprising instructions; and one or more processors in communication with the memory, wherein the one or more processors execute the instructions to perform the method of Claim 13. Sanchez Bermudez teaches a system, comprising: a non-transitory memory storage comprising instructions; and one or more processors in communication with the memory, wherein the one or more processors execute the instructions to perform the method (computer program herein may comprise instructions, the instructions comprising means for causing the above system to perform the method, the program may be implemented as a product tangibly embodied in a machine-readable storage device for execution by a programmable processor, [0104]). Thus, Claim 35 is rejected under the same rationale as Claim 13.
As per Claims 36-38, 40, 44, and 46, these claims are similar in scope to Claims 2, 3, 19, 23, 29, and 34 respectively, and therefore are rejected under the same rationale. As per Claim 47, Claim 47 is similar in scope to Claim 35, and therefore is rejected under the same rationale. As per Claims 48-50, 54, and 56, these claims are similar in scope to Claims 2, 3, 23, 29, and 34 respectively, and therefore are rejected under the same rationale.
Claim(s) 5, 6, 24, 26, 41, and 51 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Ham (US 20240265505A1).
As per Claim 5, Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 1. Sanchez Bermudez teaches the pretrained 2D sketch-to-2D image model, as discussed in the rejection for Claim 1.
However, Sanchez Bermudez and Tsai do not expressly teach that the pretrained 2D sketch-to-2D image model is a diffusion model. However, Ham teaches wherein the pretrained 2D sketch-to-2D image model is a diffusion model (guidance information includes sketch information, the sketch information includes a sketch image depicting a scene, image processing apparatus 110, via a diffusion model, predicts an output image based on the guidance information, [0032]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the pretrained 2D sketch-to-2D image model is a diffusion model as suggested by Ham. It is well-known in the art that the iterative, denoising-based approach of the diffusion model provides greater training stability.
As per Claim 6, Sanchez Bermudez and Tsai do not expressly teach that the second 2D image constrains the pretrained 2D sketch-to-2D image model during the denoising of the noisy 2D image. However, Ham teaches wherein the second 2D image constrains the pretrained 2D sketch-to-2D image model during the denoising of the noisy 2D image (conditioning network is trained based on guidance information such as spatial layout, sketch information indicate spatial layout, the conditioning network takes a noise image, intermediate noise prediction, and guidance information as input and generates modulation parameters, intermediate noise prediction and the noise modulation parameters are combined to obtain a modified intermediate noise prediction, generates a denoised image based on the modified intermediate noise prediction, the denoised image depicts a scene based on the guidance information, [0024], sketch information includes a sketch image depicting a scene, [0032]). This would be obvious for the reasons given in the rejection for Claim 5.
As per Claims 24 and 26, these claims are similar in scope to Claims 5 and 6 respectively, and therefore are rejected under the same rationale. As per Claims 41 and 51, these claims are each similar in scope to Claim 26, and therefore are rejected under the same rationale.
Claim(s) 7, 27, 42, and 52 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1), Tsai (US012254681B2), and Ham (US 20240265505A1) in view of Lee (US 20250173962A1).
As per Claim 7, Sanchez Bermudez, Tsai, and Ham are relied upon for the teachings as discussed above relative to Claim 6.
However, Sanchez Bermudez, Tsai, and Ham do not teach wherein a user provided text is further used as another control signal that constrains the pretrained 2D sketch-to-2D image model during the denoising of the noisy 2D image. However, Lee teaches wherein a user provided text is further used as another control signal that constrains the pretrained 2D sketch-to-2D image model (receiving input of a sketch and a text on a 3D object to create, acquiring a text on the 3D object, creating a 2D image from the sketch and the text, and creating a 3D object from the 2D image, Abstract). Since Ham teaches wherein the second 2D image constrains the pretrained 2D sketch-to-2D image model during the denoising of the noisy 2D image [0024, 0032], this teaching of the user provided text from Lee can be implemented into the device of Ham so that a user provided text is further used as another control signal that constrains the pretrained 2D sketch-to-2D image model during the denoising of the noisy 2D image.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez, Tsai, and Ham so that a user provided text is further used as another control signal that constrains the pretrained 2D sketch-to-2D image model during the denoising of the noisy 2D image because Lee suggests that this way, additional text input compensates for an ambiguity problem arising in sketch input, and a text prompt on the sketch is automatically generated, thereby assisting in high-quality 3D content authoring [0019].
As per Claims 27, 42, and 52, these claims are each similar in scope to Claim 7, and therefore are rejected under the same rationale.
Claim(s) 9 and 30 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Yu (US 20200074658A1).
As per Claim 9, Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 1.
However, Sanchez Bermudez and Tsai do not teach wherein the method further comprises, at the device: repeating the optimizing of the updated representation of the 3D object using a different viewpoint. However, Yu teaches wherein the method further comprises, at the device: repeating the optimizing of the updated representation of the 3D object using a different viewpoint (repeating the step 401-404 at different viewpoints, a 3D point cloud can be computed, and a 3D model of the object can be constructed, [0058]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the method further comprises, at the device: repeating the optimizing of the updated representation of the 3D object using a different viewpoint because Yu suggests that it needs more than one viewpoint in order to construct the 3D model of the object [0058].
As per Claim 30, Claim 30 is similar in scope to Claim 9, and therefore is rejected under the same rationale.
Claim(s) 10, 31, 45, and 55 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1), Tsai (US012254681B2), and Yu (US 20200074658A1) in view of Dai (US 20110158507A1).
As per Claim 10, Sanchez Bermudez, Tsai, and Yu are relied upon for the teachings as discussed above relative to Claim 9.
However, Sanchez Bermudez, Tsai, and Yu do not teach wherein the optimizing is repeated over one or more iterations until a stopping criteria is met. However, Dai teaches wherein the optimizing is repeated over one or more iterations until a stopping criteria is met (optimizing step of updating the dynamic scene geometry model based on the dynamic scene 3D model and iterating steps 104-106 until the difference between the resulted neighboring dynamic scene 3D models or between the resulted neighboring dynamic scene geometry models falls within a predetermined threshold, when the optimized result is smaller than the predetermined threshold, it may be considered that the optimized result has been converged with textures under each view angle being incorporated into the dynamic scene 3D model, so that the iteratively optimizing may be stopped, [0071]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez, Tsai, and Yu so that the optimizing is repeated over one or more iterations until a stopping criteria is met because Dai suggests that this ensures that the results are optimized to an acceptable threshold, to ensure that the 3D model is at an acceptable quality [0071].
As per Claims 31, 45, and 55, these claims are each similar in scope to Claim 10, and therefore are rejected under the same rationale.
Claim(s) 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Damrow (US 20230351347A1).
Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13.
However, Sanchez Bermudez and Tsai do not teach wherein the representation of the 3D object is a neural radiance field (NeRF) model. However, Damrow teaches wherein the representation of the 3D object is a neural radiance field (NeRF) model (NeRF (Neural Radiance Fields), is a technique that leverages machine learning to create a continuous 3D representation of a scene from a set of 2D images, [0046], in order to convert this intellectual thought into a 3D object file, the artist could first describe the envisioned sculpture in detail, the processing circuits would ten receive this description, in the form of a rough sketch, utilizing machine learning, and computer vision techniques, the processing circuits would interpret and process the artist’s input, extracting key information about the envisioned sculpture, these techniques would help the processing circuits create a digital representation of the sculpture based on the provided information, the generated 3D object file would serve as a digital twin of the artist’s envisioned sculpture, reflecting the unique characteristics described by the artist, [0101]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the representation of the 3D object is a neural radiance field (NeRF) model as suggested by Damrow. It is well-known in the art that NeRF generates highly realistic and detailed 3D scenes from 2D images, capturing complex light and shadow interactions, and producing photorealistic novel views with impressive depth and continuity.
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Ramani (US 20060114252A1).
Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13.
However, Sanchez Bermudez and Tsai do not teach wherein the representation of the 3D object is a signed distance function. However, Ramani teaches wherein the representation of the 3D object is a signed distance function (a triple S={pi|Ni,Ai,Di)} is used to represent a 3D shape, in which Di represents the signed distance, [0067], take 2D sketches and convert them to 3D models, [0128]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the representation of the 3D object is a signed distance function as suggested by Ramani. It is well-known in the art that a signed distance function enables infinite resolution without pixelation, allowing for complex shape manipulation with mathematical operators, using minimal memory by representing shapes as equations rather than meshes, and supporting advanced techniques for accurate rendering, which makes it ideal for real-time graphics, 3D modeling, and applications requiring smooth, continuous representations of geometry.
Claim(s) 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Damrow (US 20230351347A1) and Hsieh (US 20250200861A1).
Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13.
However, Sanchez Bermudez and Tsai do not teach wherein the representation of the 3D object is a Gaussian Splatting representation. However, Damrow teaches wherein the representation of the 3D object is a neural radiance field (NeRF) model [0046, 0101]. This would be obvious for the reasons given in the rejection for Claim 14.
However, Sanchez Bermudez, Tsai, and Damrow do not teach wherein the representation of the 3D object is a Gaussian Splatting representation. However, Hsieh teaches Gaussian splattering technology is a rendering technology in 3D modeling based on point cloud, which can project each point onto the image plane and use Gaussian function to render each point to generate a continuous image. Gaussian splattering technology helps to improve rendering effects in the scenario of applying NeRF, especially in the case of dealing with 3D scenes with complex geometries and details [0028]. Since Damrow teaches wherein the representation of the 3D object is a neural radiance field (NeRF) model [0046, 0101], this teaching of Gaussian Splatting from Hsieh can be implemented into the NeRF model of Damrow to improve the rendering effects of the NeRF, and thus the representation of the 3D object is a Gaussian Splatting representation.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez, Tsai, and Damrow so that the representation of the 3D object is a Gaussian Splatting representation because Hsieh suggests that this improves the rendering effects [0028].
Claim(s) 20-21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Ben-David (US 20130009950A1).
As per Claim 20, Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13. Sanchez Bermudez teaches performing an optimization of a representation of a 3D object from a 3D free-form sketch of the 3D object by: rendering the representation of the 3D object from a defined camera position [0027, 0129].
However, Sanchez Bermudez and Tsai do not teach wherein the defined camera position is selected based on the 3D free-form sketch. However, Ben-David teaches wherein the defined camera position is selected as a camera position that captures a maximum amount of information from the image (specifies as the view selection criteria the most detailed view of a portion of an area of operations, [0038], view selection criteria includes a criterion other than viewpoint, based on the view selection criteria, selecting at least one of the regions for at least one of the portions and rendering a multi-view using the selected regions in combination with the three-dimensional model of the scene, Abstract, generating a three-dimensional view of a real scene, based on criteria other than just point of view, [0001]). Since Sanchez Bermudez teaches performing an optimization of a representation of a 3D object from a 3D free-form sketch of the 3D object by: rendering the representation of the 3D object from a defined camera position [0027, 0129], this teaching from Ben-David can be implemented on the 3D free-form sketch of Sanchez Bermudez so that the defined camera position is selected as a camera position that captures a maximum amount of information from the 3D free-form sketch, and thus the defined camera position is selected based on the 3D free-form sketch.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the defined camera position is selected based on the 3D free-form sketch because Ben-David suggests that this way, the 3D model is generated with more details [0038] (Abstract) [0001].
As per Claim 21, Sanchez Bermudez teaches performing an optimization of a representation of a 3D object from a 3D free-form sketch of the 3D object by: rendering the representation of the 3D object from a defined camera position [0027, 0129].
However, Sanchez Bermudez and Tsai do not teach wherein the defined camera position is selected as a camera position that captures a maximum amount of information from the 3D free-form sketch. However, Ben-David teaches wherein the defined camera position is selected as a camera position that captures a maximum amount of information from the image (specifies as the view selection criteria the most detailed view of a portion of an area of operations, [0038], view selection criteria includes a criterion other than viewpoint, based on the view selection criteria, selecting at least one of the regions for at least one of the portions and rendering a multi-view using the selected regions in combination with the three-dimensional model of the scene, Abstract, generating a three-dimensional view of a real scene, based on criteria other than just point of view, [0001]). Since Sanchez Bermudez teaches performing an optimization of a representation of a 3D object from a 3D free-form sketch of the 3D object by: rendering the representation of the 3D object from a defined camera position [0027, 0129], this teaching from Ben-David can be implemented on the 3D free-form sketch of Sanchez Bermudez so that the defined camera position is selected as a camera position that captures a maximum amount of information from the 3D free-form sketch.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the defined camera position is selected as a camera position that captures a maximum amount of information from the 3D free-form sketch because Ben-David suggests that this way, the 3D model is generated with more details [0038] (Abstract) [0001].
Claim(s) 22 and 39 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Beltrand (US 20250078407A1).
As per Claim 22, Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13.
However, Sanchez Bermudez and Tsai do not teach wherein the representation of the 3D object is rendered from the defined camera position using a differentiable renderer. However, Beltrand teaches wherein the representation of the 3D object is rendered from the defined camera position using a differentiable renderer (modifying a three-dimensional modeled object in a 3D scene by using a user sketch, [0002], differentiable renderer, [0078]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the representation of the 3D object is rendered from the defined camera position using a differentiable renderer as suggested by Beltrand. It is well-known in the art that this allows for the optimization of 3D scene parameters to minimize the difference between a rendered image and a target image.
As per Claim 39, Claim 39 is similar in scope to Claim 22, and therefore is rejected under the same rationale.
Claim(s) 25 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Turek (US 20190317739A1).
Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13.
However, Sanchez Bermudez and Tsai do not teach wherein the pretrained 2D sketch-to-2D image model is a multi-layer perceptron (MLP). However, Turek teaches wherein the pretrained 2D sketch-to-2D image model is a multi-layer perceptron (MLP) (sketches 104 are converted to wireframes 106, [0015], multi-layer perceptron (MLP) neural network 526, [0046]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Bermudez and Tsai so that the pretrained 2D sketch-to-2D image model is a multi-layer perceptron (MLP) as suggested by Turek. It is well-known in the art that MLP is a simple and fast type of neural network.
Claim(s) 28, 43, and 53 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sanchez Bermudez (US 20200210636A1) and Tsai (US012254681B2) in view of Olson (US 20250005859A1).
As per Claim 28, Sanchez Bermudez and Tsai are relied upon for the teachings as discussed above relative to Claim 13.
However, Sanchez Bermudez and Tsai do not teach wherein the loss is a Score Distillation Sampling (SDS) loss. However, Olson teaches wherein the loss is a Score Distillation Sampling (SDS) loss [0079].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Sanchez Berumdez and Tsai so that the loss is a Score Distillation Sampling (SDS) loss as suggested by Olson. It is well-known in the art that this offers high-quality results with less direct supervision.
As per Claims 43 and 53, these claims are each similar in scope to Claim 28, and therefore are rejected under the same rationale.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 JONI HSU whose telephone number is (571)272-7785. The examiner can normally be reached M-F 10am-6:30pm.
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, Kee Tung can be reached at (571)272-7794. 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.
JH
/JONI HSU/Primary Examiner, Art Unit 2611