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 .
Claim Objections
Claim 10 is objected under 35 U.S.C. § 112 ¶2, as it recites “and/or” conjunctions.1 Examiner determined that the conjunctions cause the metes and bounds of the claim unclear as they are grammatically imprecise. If the ambiguity creates a situation where a person of ordinary skill in the art would read the language with more than one reasonable interpretation, the claim is susceptible to an indefiniteness rejection.2
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1 and 14–15 are rejected under 35 U.S.C. § 112(b) or 35 U.S.C. § 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1 and 14–15 recite the limitation “the trained model back.” (emphasis added) There is insufficient antecedent basis for this limitation in the claim.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1–15 are rejected under 35 U.S.C. § 101 because:
CLAIM 1 –
Full Text:
“A method of camera calibration for cameras deployed at monitored areas, the method comprising: obtaining a trained model, the trained model having been trained using a neural radiance field, the trained model back projects pixels in a camera image into a 3D coordinate system associated with a monitored area; and using the trained model to generate a set of calibration parameters.”
Parsed Limitations:
A method (process claim)
Of camera calibration for cameras deployed at monitored areas
Obtaining a trained model
The trained model having been trained using a neural radiance field (NeRF)
The trained model back projects pixels in a camera image into a 3D coordinate system associated with a monitored area
Using the trained model to generate a set of calibration parameters
STEP 1: STATUTORY CATEGORY DETERMINATION
Conclusion: The claim recites a process within the statutory category of 35 U.S.C. § 101(a).
Analysis:
The claim is directed to “a method” with recited steps: “obtaining,” “back projects,” and “using.”
A method claim constitutes a process, which is explicitly enumerated in § 101.
The claim does not recite a machine, manufacture, or composition of matter as the primary subject matter.
Result: Claim satisfies the statutory category requirement and proceeds to Step 2A.
STEP 2A, PRONG 1: IDENTIFY JUDICIAL EXCEPTION(S)
Conclusion: The claim recites a judicial exception under the abstract idea grouping of mathematical concepts and potentially mental processes.
Offending Clauses:
“obtaining a trained model, the trained model having been trained using a neural radiance field”
This recites training a neural network using NeRF, which is fundamentally a mathematical concept—specifically, a machine learning algorithm that learns volumetric scene functions (radiance fields) via multi-layer perceptrons (MLPs).
Per PEG § II.B.1(a), mathematical concepts include “formulas, calculations, data structures” and “methods of organizing human activity” when implemented as mental processes.
NeRF training involves mathematical operations: solving optimization problems via backpropagation, computing loss functions, and updating network weights—all mathematical abstractions.
“the trained model back projects pixels in a camera image into a 3D coordinate system”
This recites back projection, which is a mathematical transformation (inverse projection).
Back projection is a well-known mathematical operation in computer vision: mapping 2D pixel coordinates to 3D world coordinates via matrix operations.
The specification (¶41) describes this as the backward NeRF model that “for a given camera image and an approximate pose of the image, it provides a list of corresponding points in 2D coordinate system of the camera and voxels in 3D coordinate system of the facility.”
This is a mathematical calculation without technological context.
“using the trained model to generate a set of calibration parameters”
This recites applying a mathematical model to compute outputs (calibration parameters).
The specification (¶48) states: “The collected pair points (step 52), that is the 2D position of a point in the camera’s view and its corresponding 3D position in the facility’s coordinate system, may be used to calibrate the camera by calculating its calibration matrix (step 24).”
Calculating a calibration matrix is a mathematical computation (projection matrix calculation).
Citation to PEG Groupings:
Per 2019 PEG § II.B.1(a), Grouping 1 (Mathematical Concepts):
“Mathematical concepts include: (1) mathematical relationships and formulas; (2) mathematical algorithms; (3) mathematical models; (4) calculations using mathematical concepts; (5) data structures; and (6) other mathematical concepts.”
The claim recites:
A mathematical model (NeRF, a neural network learning volumetric scene functions)
Mathematical algorithms (training via backpropagation; back projection transformation)
Calculations (generating calibration parameters via matrix operations)
STEP 2A, PRONG 2: INTEGRATION INTO A PRACTICAL APPLICATION
Conclusion: The claim FAILS to integrate the judicial exception into a practical application under Step 2A, Prong 2.
Analysis:
To satisfy Step 2A, Prong 2, the claim must demonstrate that the recited judicial exception is integrated into a practical application by:
Improving a technology or technological process (e.g., computer functioning, camera functionality);
Applying the exception to a particular machine or apparatus;
Transforming an article into a different state or thing; or
Other meaningful limitations beyond merely applying the exception.
Per 2019 PEG § II.B.2(a):
“A claim that recites a judicial exception may still be directed to a patent-eligible application of that exception if the claim as a whole integrates the exception into a practical application.”
Assessment:
1. No Claimed Technological Improvement
The claim does not recite:
A specific improvement to camera functioning or calibration accuracy (e.g., “reduces calibration error by X%,” “improves calibration speed by Y factor”)
A technical mechanism that solves a technological problem beyond the abstract application of NeRF
Any hardware-specific implementation or architectural innovation
The specification (¶37) acknowledges the problem: “Existing approaches for camera calibration are labor-intensive and financially burdensome,” but the claim does not recite how the present method improves upon this—it merely states that NeRF is used.
2. No Particular Machine or Apparatus
The claim recites:
“cameras deployed at monitored areas” – a generic field-of-use limitation, not a particular machine
“a trained model” – a generic reference to a neural network, without specific architecture, hardware constraints, or technological limitations
“a 3D coordinate system” – an abstract mathematical construct, not a particular machine
The specification (¶¶43–45) describes a “computing environment” and “automated camera system,” but the claim itself does not recite these specific components or their interaction. The claim is agnostic to the underlying hardware.
3. No Transformation of an Article
While the claim recites “back projects pixels in a camera image into a 3D coordinate system,” this is:
A mathematical transformation of data (pixels → 3D coordinates)
Not a transformation of a physical article into a different state or thing
Merely the application of a mathematical formula to data
Per Alice Corp. v. CLS Bank, 573 U.S. 208 (2014), and the PEG, data manipulation alone does not constitute a practical application.
4. Insignificant Extra-Solution Activity
The steps recited are:
Obtain a trained model (obtain pre-existing mathematical construct)
Use the model to back project (apply mathematical formula)
Generate calibration parameters (output mathematical result)
These steps are routine, conventional, and insignificant in the context of applying NeRF. They represent the ordinary operation of a neural network without additional technological constraints or improvements.
Per 2019 PEG § II.B.2(b):
“Mere instructions to apply the exception, insignificant extra-solution activity, or field-of-use limitations do not suffice.”
The claim provides “mere instructions” to apply NeRF and back projection—it does not recite how this application improves camera calibration technology or solves a technological problem beyond the abstract mathematical operations.
5. Field-of-Use Limitation
The phrase “for cameras deployed at monitored areas” is a field-of-use limitation—it merely restricts the abstract mathematical operations to a particular industry or application domain (retail, logistics, etc.). Per the PEG, field-of-use limitations alone do not integrate an exception into a practical application.
Conclusion on Step 2A, Prong 2:
The claim does not integrate the recited mathematical concepts (NeRF training, back projection, calibration matrix calculation) into a practical application. The claim recites generic steps to apply well-known mathematical operations without demonstrating:
A specific technological improvement
A particular machine or apparatus
A transformation of a physical article
Meaningful limitations beyond applying the exception
STEP 2B: SIGNIFICANTLY MORE / WELL-UNDERSTOOD, ROUTINE, CONVENTIONAL (WURC)
Conclusion: The additional claim elements are NOT “significantly more” than the judicial exception because they consist of well-understood, routine, and convention (WURC) elements under Berkheimer standards.
Analysis:
Per 2019 PEG § II.B.3:
“If a claim recites a judicial exception and does not integrate the exception into a practical application under Step 2A, Prong 2, then determine whether the additional elements in the claim provide ‘significantly more’ than the exception itself. The additional elements must be more than well-understood, routine, and conventional (WURC) in the relevant field.”
Identification of Additional Elements:
Beyond the recited judicial exceptions (NeRF training, back projection, calibration parameter generation), the claim recites:
“obtaining a trained model” – obtaining a pre-existing or previously trained neural network
“cameras deployed at monitored areas” – field-of-use limitation
“a set of calibration parameters” – output of a mathematical operation
WURC Analysis:
1. Neural Radiance Field (NeRF) Training – WURC
Evidence from Specification: The specification (¶ 0057) states: “NeRF is a technique for rendering 3D scenes. NeRF has enabled the creation of high-quality visualization that are difficult or impossible to achieve using traditional rendering methods.”
NeRF was published by Mildenhall et al. in 2020 (NeRF: Representing Scenes for View Synthesis by Optimizing Radiance Fields, ECCV 2020) and is now a well-established, published technique in computer vision.
Training a neural network using NeRF via backpropagation and gradient descent is a routine and conventional machine learning operation as of the filing date (2026).
The specification does not disclose any novel modifications to NeRF training; it applies the standard NeRF algorithm.
Berkheimer Support: The applicant provides no technical evidence (e.g., performance metrics, comparative studies, implementation details) demonstrating that NeRF training is non-routine or non-conventional in the field of computer vision and machine learning.
2. Back Projection – WURC
Back projection (inverse projection) is a well-known mathematical operation in computer vision, used for decades in camera calibration, 3D reconstruction, and pose estimation.
The specification (¶41) describes backward NeRF as providing “a list of corresponding points in 2D coordinate system of the camera and voxels in 3D coordinate system of the facility”—a standard application of inverse projection matrices.
Implementing back projection via a neural network (the backward NeRF model) is a routine application of a known mathematical concept to a known problem (camera calibration).
Berkheimer Support: The applicant provides no evidence that back projection via neural networks is non-routine or non-conventional. The specification describes the backward NeRF training process (¶¶62–73) as a straightforward application of MLP training with standard loss functions (MSE loss) and optimization (Adam optimizer).
3. Calibration Matrix Calculation – WURC
Calculating a camera calibration matrix (projection matrix) from corresponding 2D–3D point pairs is a well-established, routine procedure in computer vision.
The specification (¶48) references “calculating its calibration matrix (step 24)” without disclosing any novel methodology or improvement over standard calibration matrix computation (e.g., Direct Linear Transform, bundle adjustment).
The specification does not provide technical evidence that the calibration matrix calculation is non-routine.
Berkheimer Support: Absent.
4. Obtaining a Trained Model – WURC
“Obtaining” a pre-trained or previously trained model is a routine, conventional step in applying machine learning.
The claim does not recite training the model itself (that is the judicial exception); it merely recites obtaining an already-trained model.
This is an insignificant extra-solution activity and does not provide “significantly more.”
5. Field-of-Use Limitation – WURC
Applying the exception to “cameras deployed at monitored areas” (retail stores, logistics yards, warehouses, manufacturing facilities) is a field-of-use limitation, not a technological improvement.
Per the PEG, field-of-use limitations are WURC and do not provide “significantly more.”
Conclusion on Step 2B:
The additional claim elements consist entirely of WURC operations:
Obtaining a pre-trained model (routine)
Applying NeRF training (well-established technique as of 2026)
Performing back projection (decades-old mathematical operation)
Calculating calibration matrices (routine computer vision procedure)
Field-of-use limitation (monitoring retail/logistics environments)
No factual support is provided in the specification to establish that any of these elements are non-routine or non-conventional. The specification describes standard implementations of known techniques without disclosing novel modifications, unexpected results, or technological improvements.
Therefore, the additional elements do NOT amount to “significantly more” than the judicial exception.
Similarly, claims 14–15 are rejected in the same manner as claim 1.
In re dependent claims 2–13, they are also rejected under 35 U.S.C. § 101.
Claims 2–13 do not recite any additional elements that are not well-understood, routine or conventional activity. The additional limitations3 do not integrate the mental process into practical application or add significantly more to the mental process—i.e., they do not recite any of the exemplary considerations discussed above that are indicative of an abstract idea having been integrated into a practical application.4
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1–2, 6 and 14–15 are rejected under 35 U.S.C. § 102(a)(1) as being anticipated by Qu (CN 117036612 A).
Regarding claim 1, Qu discloses a method of camera calibration for cameras deployed at monitored areas, the method comprising:
obtaining a trained model, the trained model having been trained using a neural radiance field, (Qu discloses a neural radiance field trained by a deep learning model to perform a 3D reconstruction. Qu Spec. ¶69. Neural Radiance Field: A neural radiance field is a deep learning-based model used for modeling and rendering scenes.) the trained model back projects pixels in a camera image into a 3D coordinate system associated with a monitored area; and (Per Fig. 8, Qu discloses that pixel coordinates in the 3D system are calculated applying camera’s intrinsic parameters such that modeling object 700—i.e., a monitored area—is accomplished. Ibid. ¶91. The coordinates of this pixel in the three-dimensional coordinate system can be calculated using the camera's intrinsic parameters.)
using the trained model to generate a set of calibration parameters. (Qu discloses training samples to render the neural radiation field with adjusted parameters. Ibid. ¶181. New training samples are selected from the sample set to retrain the neural radiation field network model with adjusted parameters.)
PNG
media_image1.png
418
505
media_image1.png
Greyscale
Regarding claim 14, Qu discloses a computer readable medium storing computer-executable instructions for calibrating cameras deployed at monitored areas, comprising instructions for:
obtaining a trained model, the trained model having been trained using a neural radiance field, (Qu discloses a neural radiance field trained by a deep learning model to perform a 3D reconstruction. Qu Spec. ¶69. Neural Radiance Field: A neural radiance field is a deep learning-based model used for modeling and rendering scenes.) the trained model back projects pixels in a camera image into a 3D coordinate system associated with a monitored area; and (Per Fig. 8, Qu discloses that pixel coordinates in the 3D system are calculated applying camera’s intrinsic parameters such that modeling object 700—i.e., a monitored area—is accomplished. Ibid. ¶91. The coordinates of this pixel in the three-dimensional coordinate system can be calculated using the camera's intrinsic parameters.)
using the trained model to generate a set of calibration parameters. (Qu discloses training samples to render the neural radiation field with adjusted parameters. Ibid. ¶181. New training samples are selected from the sample set to retrain the neural radiation field network model with adjusted parameters.)
Regarding claim 15, Qu discloses a computing system comprising a processor and memory, the memory storing computer executable instructions that when executed by the processor, cause the system to:
obtain a trained model, the trained model having been trained using a neural radiance field, (Qu discloses a neural radiance field trained by a deep learning model to perform a 3D reconstruction. Qu Spec. ¶69. Neural Radiance Field: A neural radiance field is a deep learning-based model used for modeling and rendering scenes.) the trained model back projects pixels in a camera image into a 3D coordinate system associated with a monitored area; and (Per Fig. 8, Qu discloses that pixel coordinates in the 3D system are calculated applying camera’s intrinsic parameters such that modeling object 700—i.e., a monitored area—is accomplished. Ibid. ¶91. The coordinates of this pixel in the three-dimensional coordinate system can be calculated using the camera's intrinsic parameters.)
use the trained model to generate a set of calibration parameters. (Qu discloses training samples to render the neural radiation field with adjusted parameters. Ibid. ¶181. New training samples are selected from the sample set to retrain the neural radiation field network model with adjusted parameters.)
Regarding claim 2, Qu discloses the method, further comprising applying the set of calibration parameters to at least one camera. (Per Fig. 8, Qu discloses that pixel coordinates in the 3D system are calculated applying camera’s intrinsic parameters such that modeling object 700—i.e., a monitored area—is accomplished. Qu Spec. ¶91. The coordinates of this pixel in the three-dimensional coordinate system can be calculated using the camera's intrinsic parameters.)
Regarding claim 6, Qu discloses the method, wherein the monitored area comprises any one of a retail store, a logistic yard, a warehouse, and a manufacturing facility (a retail store, a logistic yard, a warehouse and a manufacturing facility construed as a 3D scene). (Per Fig. 1 at S2, Qu discloses a 3D scene space where a modeled object is targeted. Qu Spec. ¶87. Acquire the camera's pose information when each image is captured and establish a 3D
scene space that encloses the modeled object;)
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.
Claims 3–5 and 11–13 are rejected under 35 U.S.C. § 103 as being unpatentable over Qu in view of Wu et al. (U.S. 12,530,840 B2).
Regarding claim 3, Qu fails to specifically disclose the method, wherein the trained model and a backward trained model are obtained by:
collecting images with known positions;
training the trained model using the neural radiance field;
obtaining approximate positions of at least one camera in the monitored area;
generating training data using the trained model for each camera; and
training the backward model for each camera, the backward model using the neural radiance field.
In related art, Wu discloses the method, wherein the trained model and a backward trained model are obtained by:
collecting images with known positions; (Per Fig. 1, Wu’s preparation module 112 discloses each image is represented by a position vector in his trained model. Wu col. 5 lines 9–21. [e]ach pixel in each image of the set of images can be represented by a position vector and a direction vector.)
training the trained model using the neural radiance field; (Per Fig. 1, Wu’s preparation module 12 trains a machine learning model for a radiance field of an object. Ibid. col. 4 line 62 – col. 5 line 8. [t]he training data preparation module 112 may be configured to generate training data for training a machine learning model to encode a radiance field of an object on the machine learning model.)
obtaining approximate positions of at least one camera in the monitored area; (Per Fig. 2, Wu’s depth map generation module 202 discloses a volume density in order to evaluate densities of points in the radiance field of the object. Ibid. col. 7 lines 14–28. The depth map generation module 202 may obtain a volume density associated with each voxel by querying a machine learning model that encoded the radiance field to output the volume density for the voxel.)
generating training data using the trained model for each camera; and (Per Fig. 1, Wu’s preparation module 12 trains a machine learning model for a radiance field of an object. Ibid. col. 4 line 62 – col. 5 line 8. [t]he training data preparation module 112 may be configured to generate training data for training a machine learning model to encode a radiance field of an object on the machine learning model.)
training the backward model (the backward model construed as a second training process) for each camera, the backward model using the neural radiance field. (Per Fig. 6, Wu’s training module 602 discloses a second training process where sample points demonstrate a surface region of the object within distance. Ibid. col. 12 lines 13–24. second training process using a plurality of [s]econd test sample points, each of the first test sample points and the second test sample points located on a plurality of rays derived from the training images, the second test sample points located within a threshold distance from a surface region of the object.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Wu into the teachings of Qu to optimize radiance field processing input images such that the NeRF reconstructs complex objects in the image. Ibid. col. 3 lines 52–57.
Regarding claim 4, Qu as modified by Wu, discloses the method, further comprising defining a spanned space for training. (Per Fig. 2, Wu’s image rendering module 204 discloses that an image is rendered in the space to composite pixels by the camera transformation and projection. Wu col. 8 lines 4–12. [a]n image may be rendered (constructed) by compositing pixels in an image space as indicated by the camera transformation and projection matrices.)
Regarding claim 5, Qu fails to specifically disclose the wherein the set of calibration parameters is generated by:
collecting pairs of points; and
calculating calibration matrices for each camera.
In related art, Wu discloses the method, wherein the set of calibration parameters is generated by:
collecting pairs of points; and (Per Fig. 2, Wu’s depth map generation module 202 discloses a volume density in order to evaluate densities of points in the radiance field of the object. Wu col. 7 lines 14–28. The depth map generation module 202 may obtain a volume density associated with each voxel by querying a machine learning model that encoded the radiance field to output the volume density for the voxel.)
calculating calibration matrices for each camera. (Per Fig. 2, Wu’s image rendering module 204 discloses that an image is rendered in the space to composite pixels by the camera transformation and projection. Ibid. col. 8 lines 4–12. [a]n image may be rendered (constructed) by compositing pixels in an image space as indicated by the camera transformation and projection matrices.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Wu into the teachings of Qu to optimize radiance field processing input images such that the NeRF reconstructs complex objects in the image. Ibid. col. 3 lines 52–57.
Regarding claim 11, Qu fails to specifically disclose the method, wherein the backward model is trained by:
determining an approximate position and angle of each camera;
using the approximate position and angle of each camera to derive a probability density function to randomly sample training data on a sampling space for that camera.
In related art, Wu discloses the method, wherein the backward model is trained by:
determining an approximate position and angle of each camera; (Per Fig. 4 at step 402, Wu discloses a view direction as a target image is obtained. Wu col. 9 lines 38–48. The viewpoint may be a point from which a target image of the object to be observed, and the view direction may be the direction along which the target image to be observed.)
using the approximate position and angle of each camera to derive a probability density function to randomly sample training data on a sampling space for that camera. (Per Fig. 2, Wu’s depth map generation module 202 discloses a volume density in order to evaluate densities of points in the radiance field of the object. Wu col. 7 lines 14–28. The depth map generation module 202 may obtain a volume density associated with each voxel by querying a machine learning model that encoded the radiance field to output the volume density for the voxel.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Wu into the teachings of Qu to optimize radiance field processing input images such that the NeRF reconstructs complex objects in the image. Ibid. col. 3 lines 52–57.
Regarding claim 12, Qu fails to specifically disclose the method, wherein the neural radiance field comprises synthesizing views by learning a volumetric scene function of 3D positions and viewing directions.
In related art, Wu discloses the method, wherein the neural radiance field comprises synthesizing views by learning a volumetric scene function of 3D positions and viewing directions. (Per Fig. 2, Wu’s image rendering module 204 discloses that an image is rendered in the space to composite pixels by the camera transformation and projection. Wu col. 8 lines 4–12. [a]n image may be rendered (constructed) by compositing pixels in an image space as indicated by the camera transformation and projection matrices.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Wu into the teachings of Qu to optimize radiance field processing input images such that the NeRF reconstructs complex objects in the image. Ibid. col. 3 lines 52–57.
Regarding claim 13, Qu as modified by Wu, discloses the method, wherein learning the volumetric scene function utilizes a multi-layer perceptron. (Per Fig. 2, Wu’s depth map generation module 202 discloses a volume density in order to evaluate densities of points in the radiance field of the object. Wu col. 7 lines 14–28. The depth map generation module 202 may obtain a volume density associated with each voxel by querying a machine learning model that encoded the radiance field to output the volume density for the voxel.)
Claims 7 and 10 are rejected in 35 U.S.C. § 103 as being unpatentable over Qu in view of Lee (KR 102161452 B2).
Regarding claim 7, Qu fails to specifically disclose the method, further comprising recalibrating by:
acquiring an image from a given camera;
determine an image validation to determine if valid; and
when valid, perform a camera change detection.
In related art, Lee discloses the method, further comprising recalibrating by:
acquiring an image from a given camera; (Lee discloses valid image frames. Lee Spec. ¶48. The present invention can extract and obtain only valid image frames (or patterns) during the pattern image processing process by using a plurality of validation patterns (211, 212) to eliminate such three-dimensional coordinate errors.)
determine an image validation to determine if valid; and (Lee’s detection unit 130 determines whether the image patterns are valid. Ibid. ¶43. When the validity of a plurality of validity verification patterns is verified, the valid pattern detection unit (130) can detect the validity of target patterns located between or around a plurality of validity verification patterns within a validity verification interval and generate a large number of point clouds from the valid target patterns.)
when valid, perform a camera change detection. (Lee’s motion validity detection device determines contrast of the valid patterns in 3D data. Ibid. ¶62. [t]he present invention can detect the validity of three-dimensional point clouds due to rapid positional movement of the scanner by utilizing the phenomenon where the contrast between pixel values at the same location of the normal validity verification pattern (511) and the inverse validity verification pattern (512) decreases (for example, meaning that the intensity of the pixel value difference (pat-inv) in FIG. 3(c) drops to ‘0’ or less).)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Lee into the teachings of Qu to provide accurate validation process while a camera detects motions. Ibid. ¶7.
Regarding claim 10, Qu fails to specifically disclose the method, further comprising using the trained backward model to detect whether a camera position and/or orientation has changed.
In related art, Lee discloses the method, further comprising using the trained backward model to detect whether a camera position and/or orientation has changed. (Lee’s motion validity detection device determines contrast of the valid patterns in 3D data. Lee Spec. ¶62. [t]he present invention can detect the validity of three-dimensional point clouds due to rapid positional movement of the scanner by utilizing the phenomenon where the contrast between pixel values at the same location of the normal validity verification pattern (511) and the inverse validity verification pattern (512) decreases (for example, meaning that the intensity of the pixel value difference (pat-inv) in FIG. 3(c) drops to ‘0’ or less).)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Lee into the teachings of Qu to provide accurate validation process while a camera detects motions. Ibid. ¶7.
Claims 8–9 are rejected under 35 U.S.C. § 103 as being unpatentable over Qu in view of Lee and further in view of Elwazer et al. (U.S. 12,664,821 B2).
Regarding claim 8, Qu as modified by Lee, discloses the claimed invention, but fails to specifically disclose the method, further comprising, when the camera position has not changed:
fine-tuning the trained model; and
fine-tuning the trained backward model.
In related art, Elwazer discloses disclose the method, further comprising, when the camera position has not changed:
fine-tuning the trained model; and (Per Fig. 16 at process 1450, Elwazer discloses fine tuning an input sequence 1456. Elwazer col. 21 line 63 – col. 22 line 12. [p]rocess 1450 fine tunes the input sequence 1456,)
fine-tuning the trained backward model. (Per Fig. 3 at step 306, Elwazer’s system 100 discloses a forward-backward algorithm to train optimal parameters. Ibid. col. 14 line 64 – col. 15 line 12. [c]ompute the observation probability 306 given each sign model P.sub.i=(O|λ.sub.i) using the forward-backward algorithm.)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Elwazer into the teachings of Qu and Lee to receive multi-dimensional data points relating to automatic body movement recognition. Ibid. col. 1 line 63 – col. 2 line 11.
Regarding claim 9, Qu as modified by Lee, discloses the claimed invention, but fails to specifically disclose the method, further comprising, when the camera position has changed:
retaining the trained backward model;
collecting pairs of points;
recalculating a calibration matrix for the given camera; and
fine tuning the trained model.
In related art, Elwazer discloses the method, further comprising, when the camera position has changed:
retaining the trained backward model; (Per Fig. 3 at step 306, Elwazer’s system 100 discloses a forward-backward algorithm to train optimal parameters. Elwazer col. 14 line 64 – col. 15 line 12. [c]ompute the observation probability 306 given each sign model P.sub.i=(O|λ.sub.i) using the forward-backward algorithm.)
collecting pairs of points; (Per Fig. 2, Elwazer’s detector module 118 recognizes different skeletal points corresponding to an object’s posture. Ibid. col. 13 lines 6–27. The transition posture detector module 118 processes the 3D data, recognizing the different skeletal points and their relation to one another, to detect the presence of the transition posture 204,)
recalculating a calibration matrix for the given camera; and (Per Fig. 9, Elwazer discloses a feature matrix 932 after decoupling the object’s movement. Ibid. col. 16 line 44 – col. 17 line 23. The decoupled movement 928 is then processed using feature extraction 930, explained in greater detail in FIG. 13, which produces a movement represented in a feature matrix 932.)
fine tuning the trained model. (Per Fig. 16 at process 1450, Elwazer discloses fine tuning an input sequence 1456. Ibid. col. 21 line 63 – col. 22 line 12. [p]rocess 1450 fine tunes the input sequence 1456,)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Elwazer into the teachings of Qu and Lee to receive multi-dimensional data points relating to automatic body movement recognition. Ibid. col. 1 line 63 – col. 2 line 11.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BENEDICT LEE whose telephone number is (571)270-0390. The examiner can normally be reached 10:00-17:00 (EST).
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, Stephen R. Koziol can be reached at (408) 918-7630. 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.
/BENEDICT E LEE/Examiner, Art Unit 2665
/Stephen R Koziol/Supervisory Patent Examiner, Art Unit 2665
1 Examiner referred to Applicant’s Spec. ¶42 and he discloses that a camera replacement algorithm is conducted in the camera’s positions and (emphasis added) orientations.
2 See MPEP § 2173.02, I. CLAIMS UNDER EXAMINATION ARE CONSTRUED DIFFERENTLY THAN PATENTED CLAIMS. During examination, after applying the broadest reasonable interpretation (BRI) consistent with the specification to the claim, if the metes and bounds of the claimed invention are not clear, the claim is indefinite and should be rejected.
3 See MPEP § 2106.05(g). Examiners should carefully consider each claim on its own merits, as well as evaluate all other relevant considerations, before making a determination of whether an element (or combination of elements) is insignificant extra-solution activity.
4 For example, per Applicant’s Spec. ¶¶47 and 53, users can set camera’s approximate position to render 2D images from a field of view—3D representation given that the extrinsic and intrinsic parameters are already provided: i.e., these settings can be considered as insignificant extra-solution activity.