DETAILED ACTION
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 11/24/2025 has been entered.
Response to Arguments
Applicant has amended claims 1, 7, 13, 19, and 25; with claims 1-30 currently pending.
Applicant's arguments filed 11/24/2025 have been fully considered. Applicant’s arguments with respect to the 35 U.S.C. 103 rejection(s) claim(s) 1,6-7, 12-13, 18-19, 24-25, and 30 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Therefore the action is made NON-FINAL.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1,6-7, 12-13, 18-19, 24-25, and 30 is/are rejected under 35 U.S.C. 103 as being unpatentable over Dundar (Dundar, Aysegul, et al. "Fine Detailed Texture Learning for 3D Meshes with Generative Models, 2022.”) in view of Remelli (Remelli E, Lukoianov A, Richter S, Guillard B, Bagautdinov T, Baque P, Fua P. Meshsdf: Differentiable iso-surface extraction. Advances in Neural Information Processing Systems. 2020;33:22468-78.).
Regarding claim 7, Dundar discloses A system comprising: one or more processors to use (Dundar Section 4.3 Model and Training Details ¶1 training on a single GPU is disclosed.) one or more neural networks to generate, based, at least in part, on one or more two-dimensional images of one or more objects, texture information and geometry information (Dundar Fig. 2 an image is input to produce a corresponding 3D model (geometry information) and texture map (texture information). Sections 2.2 3D Object Reconstruction from Single-view images and 2.3 3D Object Reconstruction form Multi-view Images explain that the input image is 2D and can be a single 2D image or multiple 2D images.) generate one or more textured 3D meshes (Dundar Fig. 1; textured 3D models are inferred (generated) using pre-trained networks. Fig. 2 show the network structure: an Encoder and two decoders in (a) along with a Generator in (b) and Discriminator in (c) can be seen which are all neural networks.) corresponding to the one or more objects based, at least in part, on the texture information and the extracted one or more surface meshes (Dundar Fig. 1; the corresponding texture information is seen on the left to the textured 3D models. Fig. 2 shows the use of a differentiable renderer that is fed the refined mesh (geometric representation). Section 2.2 3D Object Reconstruction from Single- view Images ¶1 using 3D attributes (3D mesh, texture, and lighting) in the differentiable render is disclosed.)
Dundar does not explicitly disclose that the geometric information is one or more signed distance fields; perform one or more differentiable surface extractions from the generated one or more signed distance fields to extract one or more three- dimensional (3D) surface meshes from the generated one or more signed distance fields;
Remelli, however, discloses one or more signed distance fields; (Remelli Fig. 1 (a) and Section 4.2 Single-View Reconstruction¶1-Differentiable Meshes for SVR – Found on p. 6 and Section 3.1 Deep Implicit Surface Representation – found on p. 4; single-view reconstruction is performed using a 2D picture as input. A 3D representation is created using a SDF. ) perform one or more differentiable surface extractions from the generated one or more signed distance fields to extract one or more three- dimensional (3D) surface meshes from the generated one or more signed distance fields; (Remelli Fig. 1 (a) and Section 3.2 Differentiable Iso-Surface Extraction and Section 4.2 Single-View Reconstruction¶1-Differentiable Meshes for SVR – Found on p. 6; the SDF is converted in a differentiable way to a mesh.)
It would have been obvious, before the effective filing date of the claimed invention, to one of ordinary skill in the art to modify system of Dundar with the teachings of Remelli by utilizing the signed distance field from Remelli in the creation of the 3D surface meshes in order to create a 3D model without a fixed topology while still producing a surface mesh that can be used in modeling interactions with light and physical surfaces efficiently (Remelli Section 1 Introduction ¶1-2 – found on p. 1).
Regarding claim 12, the combination of Dundar and Remelli discloses the claim limitations with regards to claim 7, as described above. They further disclose wherein the textured three-dimensional (3D) meshes further include at least one of roughness, metallic, base color, or surface normal values. (Dundar Fig. 7 and Fig. 8, it can be seen in Fig. 7 that 3D meshes show the color of the car. Fig. 8 and the caption explain that the texture also include light differences, and material differences (glass vs metal reflections).)
Regarding claims 1 and 6, they are corresponding processor comprising one or more circuits claims to claims 7 and 12 and are rejected for similar reasons.
Regarding claims 13 and 18, they are corresponding processor comprising one or more circuits claims to claims 7 and 12 and are rejected for similar reasons.
Regarding claim 19, it has similar limitations to claim 7 and the reasoning is incorporated herein.
Dundar discloses that the model is built using code from previous authors (Section 4.3 Model and Training Details). However, a memory is not explicitly disclosed. However, it would have been obvious that the method implemented on an GPU (Dundar Section 4.3 Model and Training Details ¶1 training on a single GPU is disclosed.) would have also needed A machine-readable medium having stored thereon a set of instructions. It would have been obvious before the effective filing date of the claimed invention, to one of ordinary skill in the art that the system utilizing a GPU (Dundar Section 4.3 Model and Training Details ¶1 training on a single GPU is disclosed.) would also include a memory because in order to train a model utilizing previous author’s code, they code would need to be stored in a memory and memory would also be needed to keep track of network parameters on each training cycle.
Claims 24 have corresponding limitations to claims 12 and are rejected for similar reasons.
Regarding claim 25, it has similar limitations to claim 7 and the reasoning is incorporated herein. Dundar discloses A mesh generation system, (Dundar Abstract; a method for textured 3D models is described. 3D meshes are created as mentioned in Section 1 Introduction ¶4-5.)
Dundar discloses that the model is built using code from previous authors (Section 4.3 Model and Training Details). However, a memory is not explicitly disclosed. However, it would have been obvious that the method implemented on an GPU (Dundar Section 4.3 Model and Training Details ¶1 training on a single GPU is disclosed.) would have also needed and memory for storing network parameters for the one or more first neural network. It would have been obvious before the effective filing date of the claimed invention, to one of ordinary skill in the art that the system utilizing a GPU (Dundar Section 4.3 Model and Training Details ¶1 training on a single GPU is disclosed.) would also include a memory because in order to train a model utilizing previous author’s code, they code would need to be stored in a memory and memory would also be needed to keep track of network parameters on each training cycle.
Claims 30 have corresponding limitations to claims 12 and are rejected for similar reasons.
Claim(s) 2-3, 8-9, 14-15, 20-21, 26-27 is/are rejected under 35 U.S.C. 103 as being unpatentable over Dundar (Dundar, Aysegul, et al. "Fine Detailed Texture Learning for 3D Meshes with Generative Models, 2022.”) in view of Remelli (Remelli E, Lukoianov A, Richter S, Guillard B, Bagautdinov T, Baque P, Fua P. Meshsdf: Differentiable iso-surface extraction. Advances in Neural Information Processing Systems. 2020;33:22468-78.) and Oechsle (Oechsle, Michael, et al. "Texture fields: Learning texture representations in function space." Proceedings of the IEEE/CVF international conference on computer vision. 2019.).
Regarding claim 8, the combination of Dundar and Remelli discloses the claim limitations with regards to claim 7, as described above. They further discloses wherein the one or more neural networks include a generative network having (Dundar Fig. 2 Shows the overall neural network structure.) a geometry generation branch to generate the one or more signed distance fields and (Dundar Fig. 2 and Section 3.1 Mesh Learning; 3D geometry is learned. It can be seen in Fig. 2 (a) that it results in a 3D mesh of the object – see Deform decoder to Refined Mesh. Wherein it would have been obvious to utilize the signed distance field from Remelli in the creation of the 3D surface meshes in order to create a 3D model without a fixed topology while still producing a surface mesh that can be used in modeling interactions with light and physical surfaces efficiently.).a texture generation branch to generate the texture information (Dundar Fig. 2 and Section 3.1 Mesh Learning; texture is learned. It can be seen in Fig. 2 (a) that a separate texture branch is utilized and results in 3D data – see Texture Decoder to texture map.) and wherein the generative network is further to generate the one or more textured 3D meshes using the one or more signed distance fields with the generated one or more texture information. (Dundar Fig. 7; outputs of the 3D model (meshes with texture overlays) can be seen from various viewpoints. Wherein it would have been obvious to utilize the signed distance field from Remelli in the creation of the 3D surface meshes in order to create a 3D model without a fixed topology while still producing a surface mesh that can be used in modeling interactions with light and physical surfaces efficiently.)
The combination of Dundar and Remelli does not explicitly disclose the texture information as one or more texture fields.
Oechsle, however, discloses the texture information as one or more texture fields (Oechsle Section 3.1. Texture Fields ¶1-2 and 5; texture information is embedded in to 3D information.)
It would have been obvious, before the effective filing date of the claimed invention, to one of ordinary skill in the art to modify system of the combination of Dundar and Remelli with the teachings of Oechsle by utilizing texture fields from Oechsle as the texture information in order to circumvent shape discretization and parameterization with a texture representation that is independent of the shape representation of the 3D object (Oechsle Abstract).
Regarding claim 9, the combination of Dundar, Remelli and Oechsle discloses the claim limitations with regards to claim 8, as described above. They further discloses use differentiable rendering to use the textured 3D meshes, generated by the generative network, to generate two-dimensional (2D) color image (Dundar Fig. 2 and Section 3.1 Mesh Learning ¶2; a perceptual loss is used in training. It is a difference between the input image (which was 2D) and the rendered image (which would also need to be 2D) and is seen as a color image in Fig. 2 (a).) and silhouette image pairs for the one or more objects during training of the generative network. (Dundar Fig. 2 and Section 3.1 Mesh Learning ¶3 – found on p. 4; a silhouette loss is used in training. It is a difference between the silhouette rendered and the silhouette of the input object and can be seen on the right of Fig. 2 (a).)
Regarding claims 2 and 3, they are corresponding processor comprising one or more circuits claims to claims 8 and 9 and are rejected for similar reasons.
Regarding claims 14 and 15, they are corresponding processor comprising one or more circuits claims to claims 8 and 9 and are rejected for similar reasons.
Claims 20-21 have corresponding limitations to claims 8-9 and are rejected for similar reasons.
Claims 26-27 have corresponding limitations to claims 12 and are rejected for similar reasons.
Claim(s) 4, 10, 16, 22, and 28 is/are rejected under 35 U.S.C. 103 as being unpatentable over Dundar (Dundar, Aysegul, et al. "Fine Detailed Texture Learning for 3D Meshes with Generative Models, 2022.”) in view of Remelli (Remelli E, Lukoianov A, Richter S, Guillard B, Bagautdinov T, Baque P, Fua P. Meshsdf: Differentiable iso-surface extraction. Advances in Neural Information Processing Systems. 2020;33:22468-78.), Oechsle (Oechsle, Michael, et al. "Texture fields: Learning texture representations in function space." Proceedings of the IEEE/CVF international conference on computer vision. 2019.) and Neyshabur (Neyshabur, B., S. Bhojanapalli, and A. Chakrabarti. "Stabilizing GAN training with multiple random projections. arXiv 2017." arXiv preprint arXiv:1705.07831.).
Regarding claim 10, the combination of Dundar, Remelli and Oechsle discloses the claim limitations with regards to claim 8, as described above. They further discloses analyze the 2D color image and silhouette image pairs using a pair of 2D discriminators to determine one or more loss values to be used to adjust network weights for the generative network, (Dundar Fig. 2 and Section 3.1 Mesh Learning; a perceptual loss and a silhouette loss are used in training. Both can be seen on the right of Fig. 2 (a).)
The combination of Dundar, Remelli and Oechsle disclose a single discriminator and not wherein the pair of 2D discriminators includes a discriminator for the 2D color images and a separate discriminator the silhouette images.
Neyshabur, however, discloses wherein the pair of 2D discriminators includes a discriminator for the 2D color images and a separate discriminator the silhouette images. (Neyshabur Abstract and Fig. 1; multiple discriminators, with individual types of information, are used to train a generator.)
It would have been obvious, before the effective filing date of the claimed invention, to one of ordinary skill in the art to modify system of the combination of Dundar, Remelli and Oechsle with the teachings of Neyshabur by utilizing separate discriminators as taught in Neyshabur in order to provide meaningful gradients to the generator (Neyshabur Abstract).
Regarding claim 4, it is the corresponding processor comprising one or more circuits claim to claim 10 and is rejected for similar reasons.
Regarding claims 16, it is the corresponding processor comprising one or more circuits claim to claim 10 and is rejected for similar reasons.
Claim 22 has corresponding limitations to claim 10 and is rejected for similar reasons.
Claims 28 has corresponding limitations to claim 10 and is rejected for similar reasons.
Claim(s) 5, 11, 17, 23, 29 is/are rejected under 35 U.S.C. 103 as being unpatentable over Dundar (Dundar, Aysegul, et al. "Fine Detailed Texture Learning for 3D Meshes with Generative Models, 2022.”) in view of Remelli (Remelli E, Lukoianov A, Richter S, Guillard B, Bagautdinov T, Baque P, Fua P. Meshsdf: Differentiable iso-surface extraction. Advances in Neural Information Processing Systems. 2020;33:22468-78.) and Oechsle (Oechsle, Michael, et al. "Texture fields: Learning texture representations in function space." Proceedings of the IEEE/CVF international conference on computer vision. 2019.) as evidenced by Pavllo(Pavllo, Dario, et al. "Convolutional generation of textured 3d meshes." Advances in Neural Information Processing Systems 33 (2020): 870-882.).
Regarding claim 11, the combination of Dundar, Remelli, and Oechsle discloses the claim limitations with regards to claim 8, as described above. They further discloses select one or more pairs of latent codes, from a latent space generated from the one or more two-dimensional images of the one or more objects, and provide the one or more pairs of latent codes as input to the generative network for generating the one or more textured three-dimensional (3D) meshes (Dundar Section 3.1 Mesh Learning and Fig. 2 (a); an encoder feeds the input image into the rest of the generative network. Encoders represent information in latent space and it can be seen in Fig. 2 (a) that the encoded (latent representation) is fed to the 3D geometry branch and the texture branch.)
Although the combination of Dundar, Remelli and Oechsle does not explicitly disclose wherein respective pairs of the one or more pairs of latent codes include a geometry latent code and a separate texture latent code, Dundar explains that the encoder-decoder structure from Pavllo is used and Pavllo discloses wherein respective pairs of the one or more pairs of latent codes include a geometry latent code and a separate texture latent code. (Pavllo Section 4.2 Implementation details- Conditional generation; the latent codes is split into 32D shape embedding and a 32D color embedding, meaning the color (texture) codes and shape codes are separate. Therefore the limitation is considered to be taught by Dundar as evidenced by Pavllo.)
Regarding claim 5, it is the corresponding processor comprising one or more circuits claim to claim 11 and is rejected for similar reasons.
Regarding claims 17, it is the corresponding processor comprising one or more circuits claim to claim 11 and is rejected for similar reasons.
Claim 23 has corresponding limitations to claim 11 and is rejected for similar reasons.
Claims 29 has corresponding limitations to claim 11 and is rejected for similar reasons.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MEREDITH TAYLOR whose telephone number is (571)270-5805. The examiner can normally be reached M-Th 7:30-5. Examiner’s email is Meredith.taylor@uspto.gov.
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, Vincent Rudolph can be reached on (571)272-8243. 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.
/MEREDITH TAYLOR/ Examiner, Art Unit 2671
/VINCENT RUDOLPH/ Supervisory Patent Examiner, Art Unit 2671
Prosecution Insights