CTFR 18/114,177 CTFR 93580 DETAILED ACTION Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Drawings 06-37 AIA The drawings were received on 10/24/2019 . These drawings are acceptable . Information Disclosure Statement The information disclosure statements (IDS) submitted on 06/17/2025, 02/10/2025, and 03/22/2023. Accordingly, the information disclosure statements are being considered by the examiner. Claim Rejections - 35 USC § 101 07-04-01 AIA 07-04 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 28-48 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e. an abstract idea) without significantly more. Claim 28 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. encode one or more images into equivariant latent space to produce one or more embeddings; transform the one or more embeddings by an amount of transformation to produce one or more transformed embeddings; decode the one or more transformed embeddings to produce one or more transformed versions of one or more objects depicted within the one or more images; and estimate a value for a property of the one or more transformed versions of the one or more objects within the one or more images based, at least in part, on the amount of transformation of the one or more transformed embeddings . (Considered directed to Mathematical concepts – mathematical relationships (see MPEP § 2106.04(a)(2), subsection I);) … estimate a value for a property of the one or more transformed versions of the one or more objects within the one or more images based, at least in part, on the amount of transformation of the one or more transformed embeddings . (Considered directed to a Mental Process: Making evaluations and judgements of observations for formulating observations, evaluations and judgements as claimed; see MPEP § 2106.04(a)(2), subsection IIIl)) Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). processor, comprising circuitry to use one or more neural networks to: … (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation merely include instructions to implement an abstract idea on a computer, or merely use a computer as a tool to perform an abstract idea; Thus claim limitations amount to mere instructions to apply the judicial exception using a computer/computing environment as a tool, as discussed in MPEP § 2106.05(f).) The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment and directed to invoking computers or other machinery merely as a tool to perform the claimed process/judicial exception. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Claim 29 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. wherein the one or more transformed versions of the one or more objects are derived from the one or more embeddings rotated by the amount of transformation. (Considered directed to Mathematical concepts – mathematical relationships (see MPEP § 2106.04(a)(2), subsection I);) Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Claim 30 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. wherein the value for the property of the one or more transformed versions of the one or more objects within the one or more images is based, at least in part, on the one or more transformed embeddings exhibiting a change to at least one property of the one or more objects in proportion to the amount of transformation of the one or more embeddings. (Considered directed to Mathematical concepts – mathematical relationships (see MPEP § 2106.04(a)(2), subsection I);) Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Claim 31 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. wherein the value for the property of the one or more objects within the one or more images is based, at least in part, on a comparisons between a predicted value of the property for the one or more transformed versions of the one or more objects with a known value of the property for the one or more objects . (Considered directed to Mathematical concepts – mathematical relationships (see MPEP § 2106.04(a)(2), subsection I);) Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Claim 32 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. wherein the equivariant latent space is configured such that the amount of transformation performed within the equivariant latent space is reflected in a proportional amount of change in a property of interest. (Considered directed to Mathematical concepts – mathematical relationships (see MPEP § 2106.04(a)(2), subsection I);) Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Claim 33 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. wherein the value of the property of the one or more transformed versions of the one or more objects within the one or more images is based, at least in part, on the one or more transformed embeddings containing an encoded property. (Considered directed to Mathematical concepts – mathematical relationships (see MPEP § 2106.04(a)(2), subsection I);) Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Claim 34 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. Abstract idea incorporated from claim 28. Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). wherein the one or more transformed embeddings are generate at least in part by an encoder trained to map input to points in the equivariant latent space. (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation merely include instructions to implement an abstract idea on a computer, or merely use a computer as a tool to perform an abstract idea; Thus claim limitations amount to mere instructions to apply the judicial exception using a computer/computing environment as a tool, as discussed in MPEP § 2106.05(f).) The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment and directed to invoking computers or other machinery merely as a tool to perform the claimed process/judicial exception. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. Regarding claims 35-41, the claim limitations are similar to the limitations in claims 28-34 and thus rejected under the same rationale. Regarding claim 42, the claim limitations are similar to the limitations in claim 28 and thus rejected under the same rationale. Regarding claims 43-47, the claim limitations are similar to the limitations in claims 30-34 and thus rejected under the same rationale. Claim 48 : Does claim fall within a statutory category? Yes. Step 2A Prong 1 : Evaluate whether the claim recites a judicial exception. Abstract idea incorporated from claim 47. Step 2A Prong 2 : Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h). further comprising training, using at least the value and the one or more transformed versions of the one or more objects, a neural network to estimate properties of interest. (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation merely include instructions to implement an abstract idea on a computer, or merely use a computer as a tool to perform an abstract idea; Thus claim limitations amount to mere instructions to apply the judicial exception using a computer/computing environment as a tool, as discussed in MPEP § 2106.05(f).) The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above. Step 2B : Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. The additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception; and that generally link the use of a judicial exception to a particular technological environment and directed to invoking computers or other machinery merely as a tool to perform the claimed process/judicial exception. These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B. Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible. As shown above, claims 28-48 are rejected under 35 U.S.C. 101 because the claimed invention is directed a judicial exception and does not recite, when claim elements are examined individually and as a whole, elements that the courts have identified as "significantly more” than the recited judicial exception. The claims are therefore directed to an abstract idea. Claim Rejections - 35 USC § 103 07-20-aia AIA 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. 07-21-aia AIA Claim s 28-48 are rejected under 35 U.S.C. 103 as being unpatentable over Rhodin et al. (NPL: Unsupervised Geometry-Aware Representation for 3D Human Pose Estimation, hereinafter ‘Rho’) in view of Sommer et al. (US 20210312659, hereinafter ‘So’) . Regarding independent claim 28, Rho teaches a processor, circuitry to use one or more neural networks to: (in As depicted in Fig 2: PNG media_image1.png 440 1034 media_image1.png Greyscale ) encode one or more images into equivariant latent space to produce one or more embeddings; (As depicted in Fig. 2: PNG media_image2.png 252 1022 media_image2.png Greyscale Fig. 2: Representation learning. We learn a representation that encodes geometry [ equivariant latent space ] and thereby 3D pose information in an unsupervised manner [ encode one or more images into equivariant latent space to produce one or more embeddings ]. Our method (Left) extends a conventional auto encoder (Right) with a 3D latent space, rotation operation [ equivariant latent space ], and background fusion module. The 3D rotation enforces explicit encoding of 3D information. The background fusion enables application to natural images. ) transform the one or more embeddings by an amount of transformation to produce one or more transformed embeddings; decode the one or more transformed embeddings to produce one or more transformed versions of one or more objects depicted within the one or more images; (in Pg. 5 Sec: Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints, we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j [ transform the one or more embeddings by an amount of transformation to produce one or more transformed embeddings ], as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j . However, this does not force the latent representation to encode 3D information explicitly. To this end, we model the latent representation L 3D ∈ R 3×N as a set of N points in 3D space [ transform the one or more embeddings by an amount of transformation to produce one or more transformed embeddings ] by designing the encoder E θ e and decoder D θ e so that they have a three channel output and input, respectively, as shown on the left side of Fig. 2. This enables us to model the view-change as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix [ transform the one or more embeddings by an amount of transformation to produce one or more transformed embeddings ] before using it as input to the decoder [ decode the one or more transformed embeddings to produce one or more transformed versions of one or more objects depicted within the one or more images ].) and estimate a value for a property of the one or more transformed versions of the one or more objects within the one or more images based, at least in part, on the amount of transformation of the one or more transformed embeddings. (in pgs. 5-6: … Formally, the output of the resulting autoencoder A… can be written as PNG media_image3.png 60 880 media_image3.png Greyscale and the weights θ d and θ e are optimized to minimize ||A θ I i , R i → j ) − I j || over the training set U . In this setup, which was also used in [4, 49] and is inspired by [11], the decoder D does not need to learn how to rotate the input to a new view but only how to decode the 3D latent vector L 3D [ and estimate a value for a property of the one or more transformed versions of the one or more objects within the one or more images based, at least in part, on the amount of transformation of the one or more transformed embeddings ]. This means that the encoder is forced to map to a proper 3D latent space, that is, one that can still be decoded by D after an arbitrary rotation [ based, at least in part, on the amount of transformation of the one or more transformed embeddings ]. However, while L 3D now encodes multi-view geometry, it also encodes the background and the person’s appearance. Our goal now is to isolate them from L 3D and to create two new vectors B and L app that encode the latter two so that L 3D only represents geometry and 3D pose [ based, at least in part, on the amount of transformation of the one or more transformed embeddings ]. ) While Rho teaches the information processing system as noted above and one of ordinary skill would understand that these systems operate using processing circuitry. Rho does not expressly disclose a processor, circuitry. So does expressly disclose a processor, circuitry, in [0046] As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as an apparatus, method or computer program product. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system.” [ a processor, circuitry to use one or more neural networks ] Furthermore, aspects of the present invention may take the form of a computer program product embodied in one or more computer readable medium(s) having computer executable code embodied thereon… [0050] A ‘processor’ as used herein encompasses an electronic component which is able to execute a program or machine executable instruction or computer executable code. References to the computing device comprising “a processor” should be interpreted as possibly containing more than one processor or processing core… And in [0029] The encoder neural network and the generator neural network may be two separate neural networks. They could also be incorporated into a single neural network which is similar to an auto encoder convolutional neural network which is typically used for image processing or image modification [ a processor, circuitry to use one or more neural networks ]… So and Rho are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art for retrieving information from and processing image data using machine learning models, as disclosed by So with the method of developing information retrieval and processing techniques using machine learning systems and algorithms, as disclosed by Rho. One of ordinary skill in the arts would have been motivated to combine the methods disclosed by So and Rho, as noted above. Doing so allows for processing image augmentation transformations using machine learning techniques and models, (So, 0093). Regarding claim 29, the rejection of claim 28 is incorporated and Rho in combination with So further teaches the processor of claim 28, wherein the one or more transformed versions of the one or more objects are derived from the one or more embeddings rotated by the amount of transformation . (in Pgs. 4-6 Sec. Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints, we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j [ wherein the one or more transformed versions of the one or more objects are derived from the one or more embeddings rotated by the amount of transformation ]… This enables us to model the view-change as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix before using it as input to the decoder… In this setup, which was also used in [4, 49] and is inspired by [11], the decoder D does not need to learn how to rotate the input to a new view but only how to decode the 3D latent vector L³ D . This means that the encoder is forced to map to a proper 3D latent space, that is, one that can still be decoded…pose changes, this results in L encoding pose while L encodes appearance… In practice, the encoder E has two outputs, that is, Eθ : Iⁱ → (L³D, Lᵃᵖᵖ) and the decoder Dθd accepts these plus the background as inputs, after swapping appearance and rotating the geometric representation for two views i and j [ wherein the one or more transformed versions of the one or more objects are derived from the one or more embeddings rotated by the amount of transformation ]. We therefore write the output of our encoder-decoder as PNG media_image4.png 44 886 media_image4.png Greyscale …) Regarding claim 30, the rejection of claim 29 is incorporated and Rho in combination with So further teaches the processor of claim 29, wherein the value for the property of the one or more transformed versions of the one or more objects within the one or more images is based, at least in part, on the one or more transformed embeddings exhibiting a change to at least one property of the one or more objects in proportion to the amount of transformation of the one or more embeddings . (in Pgs. 4-6 Sec. Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints, we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j [ based, at least in part, on the one or more transformed embeddings exhibiting a change to at least one property of the one or more objects in proportion to the amount of transformation of the one or more embeddings ]… This enables us to model the view-change [ wherein the value for the property of the one or more transformed versions of the one or more objects within the one or more images is based ] as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix [ based, at least in part, on the one or more transformed embeddings exhibiting a change to at least one property of the one or more objects in proportion to the amount of transformation of the one or more embeddings ] before using it as input to the decoder… In this setup, which was also used in [4, 49] and is inspired by [11], the decoder D does not need to learn how to rotate the input to a new view but only how to decode the 3D latent vector L³ D . This means that the encoder is forced to map to a proper 3D latent space, that is, one that can still be decoded…pose changes, this results in L encoding pose while L encodes appearance… In practice, the encoder E has two outputs, that is, Eθ : Iⁱ → (L³D, Lᵃᵖᵖ) and the decoder Dθd accepts these plus the background as inputs, after swapping appearance and rotating the geometric representation for two views i and j [ wherein the value for the property of the one or more transformed versions of the one or more objects within the one or more images is based, at least in part, on the one or more transformed embeddings exhibiting a change to at least one property of the one or more objects in proportion to the amount of transformation of the one or more embeddings ]. We therefore write the output of our encoder-decoder as PNG media_image4.png 44 886 media_image4.png Greyscale …) Regarding claim 31, the rejection of claim 28 is incorporated and Rho in combination with So further teaches the processor of claim 28, wherein the value for the property of the one or more objects within the one or more images is based, at least in part, on a comparisons between a predicted value of the property for the one or more transformed versions of the one or more objects with a known value of the property for the one or more objects . (in pg. 7: Combined optimization. To train A with sequences featuring several people and backgrounds, we randomly select mini-batches of Z triplets (Iⁱ, Iʲ, Iᵏ ) in U , with t /= t′, from individual sequences. In other words, all three views feature the same person. The first two are taken at the same time but from different view-points. The third is taken at a different time and from an arbitrary viewpoint k. For each such mini-batch, we compute the loss [ wherein the value for the property of the one or more objects within the one or more images is based, at least in part, on a comparisons between a predicted value of the property for the one or more transformed versions of the one or more objects with a known value of the property for the one or more objects ]: PNG media_image5.png 118 876 media_image5.png Greyscale where L ′ = (L³D , Lᵃᵖᵖ) is the output of encoder E applied to image Iᵏ , B is the background in view j, and Rⁱ→ʲ denotes the rotation from view i to view j [ …, at least in part, on a comparisons between a predicted value of the property for the one or more transformed versions of the one or more objects with a known value of the property for the one or more objects ]. Note that we apply E twice, to obtain L³D and Lᵃᵖᵖ in Eq. 3 while ignoring Lapp and L3D with the swap discussed above. At training time, we minimize a total loss that is the sum of the pixel-wise error Eθd ,θe of Eq. 3 [ wherein the value for the property of the one or more objects within the one or more images is based, at least in part, on a comparisons between a predicted value of the property for the one or more transformed versions of the one or more objects with a known value of the property for the one or more objects ] and a second term obtained by first applying a Resnet with 18 layers trained on ImageNet on the output and target image and then computing the feature difference [ based, at least in part, on a comparisons between a predicted value of the property for the one or more transformed versions of the one or more objects with a known value of the property for the one or more objects ] after the second block level, as previously done with VGG by [23]…) Regarding claim 32, the rejection of claim 28 is incorporated and Rho in combination with So further teaches the processor of claim 28, wherein the equivariant latent space is configured such that the amount of transformation performed within the equivariant latent space is reflected in a proportional amount of change in a property of interest . (As depicted in Fig. 2: PNG media_image2.png 252 1022 media_image2.png Greyscale Fig. 2: Representation learning. We learn a representation that encodes geometry [ wherein the equivariant latent space is configured such that the amount of transformation performed within the equivariant latent space is reflected in a proportional amount of change in a property of interest ] and thereby 3D pose information in an unsupervised manner. Our method (Left) extends a conventional auto encoder (Right) with a 3D latent space, rotation operation [ wherein the equivariant latent space is configured such that the amount of transformation performed within the equivariant latent space is reflected in a proportional amount of change in a property of interest ], and background fusion module. The 3D rotation enforces explicit encoding of 3D information. The background fusion enables application to natural images. And in Pg. 5 Sec: Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints [ wherein the equivariant latent space is configured such that the amount of transformation performed within the equivariant latent space is reflected in a proportional amount of change in a property of interest ], we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j . However, this does not force the latent representation to encode 3D information explicitly. To this end, we model the latent representation L 3D ∈ R 3×N as a set of N points in 3D space by designing the encoder E θ e and decoder D θ e so that they have a three channel output and input, respectively, as shown on the left side of Fig. 2. This enables us to model the view-change [ … the equivariant latent space is reflected in a proportional amount of change in a property of interest ] as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix [ wherein the equivariant latent space is configured such that the amount of transformation performed within the equivariant latent space is reflected in a proportional amount of change in a property of interest ] before using it as input to the decoder.) Regarding claim 33, the rejection of claim 28 is incorporated and Rho in combination with So further teaches the processor of claim 28, wherein the value of the property of the one or more transformed versions of the one or more objects within the one or more images is based, at least in part, on the one or more transformed embeddings containing an encoded property. (in Pgs. 4-6 Sec. Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints [ wherein the value of the property of the one or more transformed versions of the one or more objects within the one or more image ], we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j [ s is based, at least in part, on the one or more transformed embeddings containing an encoded property ]… This enables us to model the view-change as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix before using it as input to the decoder… In this setup, which was also used in [4, 49] and is inspired by [11], the decoder D does not need to learn how to rotate the input to a new view but only how to decode the 3D latent vector L³ D . This means that the encoder is forced to map to a proper 3D latent space, that is, one that can still be decoded…pose changes, this results in L encoding pose while L encodes appearance… In practice, the encoder E has two outputs, that is, Eθ : Iⁱ → (L³D, Lᵃᵖᵖ) and the decoder Dθd accepts these plus the background as inputs, after swapping appearance and rotating the geometric representation for two views i and j [ wherein the value of the property of the one or more transformed versions of the one or more objects within the one or more images is based, at least in part, on the one or more transformed embeddings containing an encoded property ]. We therefore write the output of our encoder-decoder as PNG media_image4.png 44 886 media_image4.png Greyscale …) Regarding claim 34, the rejection of claim 28 is incorporated and Rho in combination with So further teaches the processor of claim 28, wherein the one or more transformed embeddings are generated at least in part by an encoder trained to map input to points in the equivariant latent space . (in Pgs. 4-6 Sec. Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints, we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j [ wherein the one or more transformed embeddings are generated at least in part by an encoder trained to map input to points in the equivariant latent space ]… This enables us to model the view-change as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix [ wherein the one or more transformed embeddings are generated at least in part by an encoder trained to map input to points in the equivariant latent space ] before using it as input to the decoder… This means that the encoder is forced to map to a proper 3D latent space [ wherein the one or more transformed embeddings are generated at least in part by an encoder trained to map input to points in the equivariant latent space ], that is, one that can still be decoded…pose changes, this results in L encoding pose while L encodes appearance… In practice, the encoder E has two outputs, that is, Eθ : Iⁱ → (L³D, Lᵃᵖᵖ) and the decoder Dθd accepts these plus the background as inputs, after swapping appearance and rotating the geometric representation for two views i and j. We therefore write the output of our encoder-decoder as PNG media_image4.png 44 886 media_image4.png Greyscale …) Regarding claims 35-38, the claim limitations are similar to the limitations in claims 28-31and thus rejected under the same rationale. Regarding claim 39, the rejection of claim 35 is incorporated and Rho in combination with So further teaches the system of claim 35, wherein the equivariant latent space has the characteristic that an object can be rotated within the equivariant latest space such that a property of interest changes in proportion to an amount of rotation. (As depicted in Fig. 2: PNG media_image2.png 252 1022 media_image2.png Greyscale Fig. 2: Representation learning. We learn a representation that encodes geometry [ wherein the equivariant latent space has the characteristic that an object can be rotated within the equivariant latest space such that a property of interest changes in proportion to an amount of rotation.t ] and thereby 3D pose information in an unsupervised manner. Our method (Left) extends a conventional auto encoder (Right) with a 3D latent space, rotation operation [ wherein the equivariant latent space has the characteristic that an object can be rotated within the equivariant latest space such that a property of interest changes in proportion to an amount of rotation. ], and background fusion module. The 3D rotation enforces explicit encoding of 3D information. The background fusion enables application to natural images. And in Pg. 5 Sec: Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints [ wherein the equivariant latent space has the characteristic that an object can be rotated within the equivariant latest space such that a property of interest changes in proportion to an amount of rotation ], we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j . However, this does not force the latent representation to encode 3D information explicitly. To this end, we model the latent representation L 3D ∈ R 3×N as a set of N points in 3D space by designing the encoder E θ e and decoder D θ e so that they have a three channel output and input, respectively, as shown on the left side of Fig. 2. This enables us to model the view-change [ …the equivariant latest space such that a property of interest changes in proportion to an amount of rotation ] as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix [ wherein the equivariant latent space has the characteristic that an object can be rotated within the equivariant latest space such that a property of interest changes in proportion to an amount of rotation ] before using it as input to the decoder.) Regarding claims 40-41, the claim limitations are similar to the limitations in claims 33-34and thus rejected under the same rationale. Regarding claim 42, the claim limitations are similar to the limitations in claim 28 and thus rejected under the same rationale. Regarding claims 43-47, the claim limitations are similar to the limitations in claims 30-31, 39, 33-34 and thus rejected under the same rationale. Regarding claim 48, the rejection of claim 42 is incorporated and Rho in combination with So further teaches the method of claim 42, further comprising training, using at least the value and the one or more transformed versions of the one or more objects, a neural network to estimate properties of interest . (in Pgs. 4-6 Sec. Learning to encode multi-view geometry… To leverage multi-view geometry, we take our inspiration from Novel View Synthesis methods [36, 37, 4, 49, 11] that rely on training encoder-decoders on multiple views of the same object, such as a car or a chair. Let (I i , I j ) ∈ U be two images taken from different viewpoints but at the same time t. Since we are given the rotation matrix R i→j connecting the two viewpoints, we could feed this information as an additional input to the encoder and decoder and train them to encode I i and resynthesize I j , as in [36, 37]. Then, novel views of the object could be rendered by varying the rotation parameter R i→j [ further comprising training, using at least the value and the one or more transformed versions of the one or more objects, a neural network to estimate properties of interest ]… This enables us to model the view-change as a proper 3D rotation by matrix multiplication of the encoder output by the rotation matrix [ further comprising training, using at least the value and the one or more transformed versions of the one or more objects, a neural network to estimate properties of interest ] before using it as input to the decoder… This means that the encoder is forced to map to a proper 3D latent space [ further comprising training, using at least the value and the one or more transformed versions of the one or more objects, a neural network to estimate properties of interest ], that is, one that can still be decoded…pose changes, this results in L encoding pose while L encodes appearance… In practice, the encoder E has two outputs, that is, Eθ : Iⁱ → (L³D, Lᵃᵖᵖ) and the decoder Dθd accepts these plus the background as inputs, after swapping appearance and rotating the geometric representation for two views i and j. We therefore write the output of our encoder-decoder as PNG media_image4.png 44 886 media_image4.png Greyscale …) Response to Arguments Applicant's arguments filed 02/05/2026 have been fully considered. The applicant’s remarks are directed to amended subject matter that have not been previously examined by the examiner and the rejections addressing these amendments are noted above. See full rejection noted above. Conclusion 07-40 AIA 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. 07-96 AIA The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Sitzmann et al. (NPL: DeepVoxels: Learning Persistent 3D Feature Embeddings): teaches in section: Deep Learning for View Synthesis: A layered scene representation [56] can be learned based on a re-rendering loss. A large corpus of work focuses on embedding 2D views of scenes into a learned low-dimensional latent space that is then decoded into a novel view [54, 61, 7, 9, 60, 45, 6]. Some of these approaches rely on embedding views into a latent space that does not enforce any geometrical constraints [54, 7, 9], others enforce geometric constraints in varying degrees [60, 45, 6, 10], such as learning rotation equivariant features by explicitly rotating the latent space feature vectors. We focus on optimizing a scene-specific embedding over a training corpus of 2D observations and explicitly account for concepts from 3D vision such as perspective projection and occlusion to constrain the latent space… Worrall et al. (NPL: Interpretable Transformations with Encoder-Decoder Networks ): teaches Definition 1 A function f : X → Y is equivariant [49] under a set of transformations Θ if for any transformation T :Θ×X→X of the input, we can associate a transformation F:Θ×Y→Yof the output such that Fθ[f(x)]=f(Tθ[x]), (1) for all θ ∈ Θ. Transformations Tθ and Fθ represent the same underlying transformation but in different spaces, denoted θ. Equivariance is desirable, because it reveals to us a direct relationship between image-space and feature-space transformations, which for deep neural networks are usually elusive [31]. Note that invariance is a special case of equivariance, where Fθ=I is the identity for all input transformations. Cohen et al. (NPL: Transformation Properties of Learned Visual Representations): teaches mathematical techniques for modeling of rotating objects that employs latent representations. Kawachi et al. (US 20210326728): teaches the use of a rotation matrix in processing latent space of an encoder model for capturing and modeling information using machine learning models. Xie et al. (US 20180060704): teaches rotated images that can be recognized using an image character recognition model, the image character recognition model generated by training a set neural network model by using a rotated line character image training sample, the rotated line character training sample comprising a rotated line character image and an expected character recognition result corresponding to the rotated line character image, character units in the rotated line character image rotated by 90 degrees in relation to character units in a standard line character image. Kriegman et al. (US 20190311227): teaches utilizing an orientation neural network to determine the orientation of documents depicted in digital images the digital image character recognition system can reduce processing power required to generate searchable text. Additionally, utilizing the orientation of the document depicted in the digital image, the digital image character recognition system can avoid analyzing word boxes in multiple orientations (e.g., in all four rotations), thus requiring a fraction of the computer processing power. Similarly, the digital image character recognition system can avoid time and computational costs associated with generating training data for one or more neural networks. Hara et al. (NPL: “Designing Deep Convolutional Neural Networks for Continuous Object Orientation Estimation”): teaches the design of the orientation prediction unit. For the continuous orientation estimation task, the network has to predict the angular value, which is in a non-Euclidean space, prohibiting the direct use of a typical L2 loss function. To handle this problem, we propose three different approaches. The first approach represents an orientation as a 2D point on a unit circle, then trains the network using the L2 loss. In test time, the network’s output, a 2D point not necessarily on a unit circle, is converted back to the angular value by atan2 function. Our second approach also uses a-point-on-a-unit-circle representation, however, instead of the L2 loss, it minimizes a loss defined directly on the angular difference. Our third approach, which is significantly different from the first two approaches, is based on the idea of converting the continuous orientation estimation task into a set of discrete orientation estimation tasks and addressing each discrete orientation estimation task by a standard softmax function. In test time, the discrete orientation outputs are converted back to the continuous orientation using a mean-shift algorithm. The discretized orientations are determined such that all the discretized orientations are uniformly distributed in the output circular space. The mean-shift algorithm for the circular space is carried out to find the most plausible orientation while taking into account the softmax probability for each discrete orientation. Any inquiry concerning this communication or earlier communications from the examiner should be directed to OLUWATOSIN ALABI whose telephone number is (571)272-0516. The examiner can normally be reached Monday-Friday, 8:00am-5:00pm 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, Michael Huntley can be reached at (303) 297-4307. 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. /OLUWATOSIN ALABI/Primary Examiner, Art Unit 2129 Application/Control Number: 18/114,177 Page 2 Art Unit: 2129 Application/Control Number: 18/114,177 Page 3 Art Unit: 2129 Application/Control Number: 18/114,177 Page 4 Art Unit: 2129 Application/Control Number: 18/114,177 Page 5 Art Unit: 2129 Application/Control Number: 18/114,177 Page 6 Art Unit: 2129 Application/Control Number: 18/114,177 Page 7 Art Unit: 2129 Application/Control Number: 18/114,177 Page 8 Art Unit: 2129 Application/Control Number: 18/114,177 Page 9 Art Unit: 2129 Application/Control Number: 18/114,177 Page 10 Art Unit: 2129 Application/Control Number: 18/114,177 Page 11 Art Unit: 2129 Application/Control Number: 18/114,177 Page 12 Art Unit: 2129 Application/Control Number: 18/114,177 Page 13 Art Unit: 2129 Application/Control Number: 18/114,177 Page 14 Art Unit: 2129 Application/Control Number: 18/114,177 Page 15 Art Unit: 2129 Application/Control Number: 18/114,177 Page 16 Art Unit: 2129 Application/Control Number: 18/114,177 Page 17 Art Unit: 2129 Application/Control Number: 18/114,177 Page 18 Art Unit: 2129 Application/Control Number: 18/114,177 Page 19 Art Unit: 2129 Application/Control Number: 18/114,177 Page 20 Art Unit: 2129 Application/Control Number: 18/114,177 Page 21 Art Unit: 2129 Application/Control Number: 18/114,177 Page 22 Art Unit: 2129 Application/Control Number: 18/114,177 Page 23 Art Unit: 2129 Application/Control Number: 18/114,177 Page 24 Art Unit: 2129 Application/Control Number: 18/114,177 Page 25 Art Unit: 2129 Application/Control Number: 18/114,177 Page 26 Art Unit: 2129 Application/Control Number: 18/114,177 Page 27 Art Unit: 2129 Application/Control Number: 18/114,177 Page 28 Art Unit: 2129 Application/Control Number: 18/114,177 Page 29 Art Unit: 2129
Prosecution Insights