DETAILED ACTION
Claims 1-20 are presented for examination.
This office action is in response to submission of application on 21-JULY-2026.
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 13-JUNE-2022 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
The information disclosure statement (IDS) submitted on 03-OCTOBER-2023 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Amendment
The amendment filed 21-JULY-2026 in response to the previous office action mailed 22-APRIL-2026 has been entered. Claims 1-20 remain pending in the application.
With regards to the non-final office action’s rejection as double patenting, the rejection is held in abeyance until such time that the claims contain otherwise allowable subject matter. The previous rejection for claims 1, 3-5, 9-11, 13-15, and 20 is included within in order to have it on record.
With regards to the non-final office action’s rejections under 103, the amendments to the claims necessitated a new consideration of the art. After this consideration, the examiner respectfully disagrees with the applicant’s arguments that the art referenced in the previous office action does not teach the amendment claim limitations. A new 103 rejection over the prior art has been provided.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is ap11propriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claims 1, 3-5, 9-11, 13-15, and 20 rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1, 5-9, and 17 of U.S. Patent No. 11869149, hereinafter 11869149. Although the claims at issue are not identical, they are not patentably distinct from each other because:
Regarding claim 1:
Regarding the limitation a computer-implemented method for training a machine learning model to generate representations of point clouds, the method comprising: executing a first neural network on a first point cloud that represents a first three- dimensional (3D) scene to generate a plurality of segmentations:
11869149 contains nearly identical language in claim 1:
A computer-implemented method for training a machine learning model to generate representations of point clouds, the method comprising: executing a first neural network on a first point cloud that represents a first three- dimensional (3D) scene to generate a key set and a value set; generating an output vector set based on a first query set, the key set, and the value set
The language is identical up until the generation of the key set and value set. However, these would be equivalent to the segmentations of the present application as they are categories that the point cloud map to, which would be segmentations. This mapping is described in Paragraph 116 of both 11869149’s and the present application’s specification.
Regarding the limitation computing a plurality of spatial features based on the plurality of segmentations; computing a plurality of quantized context features based on the plurality of segmentations and a first set of codes representing a first set of 3D geometry blocks; and training the first neural network based on a likelihood of reconstructing the first point cloud, the plurality of quantized context features, and the plurality of spatial features to generate a trained neural network:
11869149 contains identical language in claim 1.
Regarding the limitation wherein a trained machine learning model includes the trained neural network and a second set of codes representing a second set of 3D geometry blocks and maps a point cloud representing a 3D scene to a representation of a plurality of 3D geometry instances:
11869149 contains nearly identical language in claim 1:
wherein a trained machine learning model includes the updated neural network, a second query set, and a second set of codes representing a second set of 3D geometry blocks and maps a point cloud representing a 3D scene to a representation of a plurality of 3D geometry instances
11869149 contains a second query set that is not seen in the present application. However, as this limitation of 11869149 contains all limitations of its counterpart within the present application, it anticipates this limitation.
Regarding claim 3, which depends upon claim 1:
Claim 3 of the present application is anticipated by claim 5 of 11869149:
Claim 5 of 11869149 recites a first spatial feature that specifies at least one of a weight. This fulfills the limitation of claim 3 in the present application that a first spatial feature is computed by computing at least one of … a weight as designating the weight would involve calculating it.
As such this claim is rejected as nonstatutory double patenting.
Regarding claim 4, which depends upon claim 1:
Claim 4 of the present application corresponds to claim 6 of 11869149:
Regarding the segmentations of claim 4, these were previously determined to correspond to the vector sets of 11869149 as discussed above in claim 1. Therefore, both recite the same invention.
As such this claim is rejected as nonstatutory double patenting.
Regarding claim 5, which depends upon claim 1:
Claim 5 of the present application corresponds to claim 17 of 11869149, as both recite the same invention. As such this claim is rejected as nonstatutory double patenting.
Regarding claim 9, which depends upon claim 1:
Claim 9 of the present application corresponds to claim 9 of 11869149, as both recite the same invention. As such this claim is rejected as nonstatutory double patenting.
Regarding claim 10, which depends upon claim 9:
Claim 10 of the present application corresponds to claim 9 of 11869149:
Claim 10 of the present application describes a first 3D geometry instance included in the first plurality of 3D geometry instances comprises a first instance of a first 3D geometry block included in the second set of 3D geometry blocks and Claim 9 of 11869149 describes at least one instance of a first 3D geometry block included in the second set of 3D geometry blocks, which describe the same invention.
Furthermore, claim 10 describes a second 3D geometry instance included in the first plurality of 3D geometry instances comprises a second instance of the first 3D geometry block or a first instance of a second 3D geometry block included in the second set of 3D geometry blocks. Claim 9 of 11869149 describes one instance of a second 3D geometry block included in the second set of 3D geometry block which would correspond to the second version of the second 3D geometry instance listed in claim 10 of the present application.
Therefore, both recite the same invention.
As such this claim is rejected as nonstatutory double patenting.
Claims 11, 13, and 15 recite a non-transitory computer readable media that parallels the method of claims 1, 3, and 5 respectively. Therefore, the analysis discussed above with respect to claims 1, 3, and 5 also applies to claims 11, 13, and 15 respectively. Accordingly, claims 11, 13, and 15 are rejected based on substantially the same rationale as set forth above with respect to claims 1, 3, and 5 respectively.
Regarding claim 14, which depends upon claim 11:
Claim 14 of the present application is anticipated by claim 6 of 11869149. Regarding the limitation of claim 14:
Claim 14 of the present application and claim 6 of 11869149 are both directed towards computing quantized context features, with the single computed quantized context feature of 11869149’s claim 6 anticipating the plurality of claim 14 of the present applications. Furthermore, claim 6 of 11869149 teaches computing context features based on the vector sets that have previously been determined to be the counterparts of the segmentations of the present application, and therefore anticipate claim 14’s computation of the context features based on the plurality of segmentations. Finally, claim 6 of 11869149 teaches computing a set of distances between a first context feature and a first set of codes, then setting the quantized context feature equal to the first code of the set based on the distances, which would anticipate claim 6’s mapping the plurality of context features to the plurality of quantized context features based on the first set of codes as it is a form of mapping.
Claims 20 recites a system that parallels the method of claim 1. Therefore, the analysis discussed above with respect to claim 1 also applies to claim 20. Accordingly, claim 20 is rejected based on substantially the same rationale as set forth above with respect to claim 1.
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.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Le et al. (Pub. No. US 20220292765 A1, filed March 15th 2021, hereinafter Le) in view of Zhao et al. (Pub. No. US 20230222628 A1, filed January 1st 2022, hereinafter Zhao).
Regarding claim 1:
Claim 1 recites:
A computer-implemented method for training a machine learning model to generate representations of point clouds, the method comprising: executing a first neural network on a first point cloud that represents a first three- dimensional (3D) scene to generate a plurality of segmentations; computing a plurality of spatial features based on the plurality of segmentations; computing a plurality of quantized context features based on the plurality of segmentations and a first set of codes representing a first set of 3D geometry blocks; and performing one or more training operations to update the first neural network using a plurality of loss terms and based on a reconstruction of the first point cloud from the plurality of quantized context features and the plurality of spatial features, wherein at least one weight associated with the plurality of loss terms varies during the one or more training operations, wherein a trained machine learning model includes the trained neural network and a second set of codes representing a second set of 3D geometry blocks and maps a point cloud representing a 3D scene to a representation of a plurality of 3D geometry instances.
Le discloses training a machine learning model to generate representations of point clouds, the method comprising: executing a first neural network on a first point cloud that represents a first three- dimensional (3D) scene to generate a plurality of segmentations:
Le teaches the prediction of a primitive set to represent to 3D point cloud (Paragraph 19) where the primitives are said to be equivalent to segmentations (Paragraph 42). Therefore Le generates a plurality of segmentations. Furthermore, this process is performed by the local primitive fitting network, which comprises one or more neural networks that are trained (Paragraph 28), which would training a machine learning model. The prediction of the primitive set (i.e., the generation of the segmentations) is therefore the execution of the neural network that the local primitive fitting network comprises.
Le discloses computing a plurality of spatial features based on the plurality of segmentations:
Le teaches obtaining global contextual information as additional input for the local primitive fitting network such as a latent vector representing a seed point for a local patch (Paragraph 16) wherein a seed point is defined as the local centroid of the patch (Paragraph 38). The global contextual information example of spatial features based on the plurality of segmentations as a patch would be analogous to a segmentation as both represent a particular identified area from the full 3D point cloud, and global contextual information would be analogous to the spatial features as they may describe particular feature related the segment’s position and shape of the local patch (or segmentation) as the centroid describes the arithmetic mean position of all points in the segmentation.
Le discloses performing one or more training operations to update the first neural network using a plurality of loss terms] and based on a reconstruction of the first point cloud from the plurality of [quantized context features] and the plurality of spatial features:
Le teaches the likelihood of reconstructing a first point cloud as it teaches the computation of a probability that each individual primitive is a particular primitive type, which is used to determine the predicted primitive type by choosing the primitive type with the highest probability (Paragraph 50). As primitives are used to construct a 3D model which is a representation (i.e., reconstruction) of the 3D point cloud (Paragraph 26), this would increase the likelihood of reconstructing the first point cloud as it is necessary that the primitive types be accurate in order to represent the point cloud. This information is then used to train a first neural network to generate a trained neural network through a primitive type loss function, therefore having the training be based on the reconstruction of the point cloud. This is part of the training process and is a calculation of the actual primitive types versus the predicted primitive types (Paragraph 33, equation 2) which updates the neural network (Paragraph 39).
Furthermore, Le teaches that the global contextual information that is analogous to the spatial features as described previously in the 103 rejection is the basis of the representation of local 3D primitives that approximate a local structure of the point cloud, or in other terms the first point cloud is reconstructed from the global contextual information, i.e. the spatial features. (Le, Paragraph 62).
However, Le does not teach the plurality of quantized context features, which are computed further below by Zhao. The quantized context features of Zhao could therefore be combined with the other features that are used to train the neural network of Le in order to produce a trained neural network that increases a likelihood that the trained neural network reconstructs the first point cloud from the plurality of quantized context features and the plurality of spatial features.
Furthermore, Le does not teach updating the first neural network using a plurality of loss terms. This is taught further below by Zhao, and may be combined with Le in order to incorporate the specific loss function usage of Zhao within the training method of Le.
Le discloses wherein a trained machine learning model includes the trained neural network and a second set of codes representing a second set of 3D geometry blocks and maps a point cloud representing a 3D scene to a representation of a plurality of 3D geometry instances:
Le teaches fitting 3D primitives, or geometry instances, to a 3D point cloud that represents a 3D scene using a fitting tool which would be a machine learning model (Paragraph 27). The primitive types would be the second set of codes that represent the 3D geometry blocks, as geometry blocks would be analogous to a variety of 3D shapes that are used to construct a 3D scene in the same manner that the primitive types are.
Regarding the second set of 3D geometry block, the first set is taught below by Zhao.
However, Le does not disclose each and every limitation of claim 1. Instead, further limitations are disclosed by Zhao.
Zhao discloses computing a plurality of quantized context features based on the plurality of segmentations and a first set of codes representing a first set of 3D geometry blocks:
Zhao in the same field of endeavor of reinforcement learning teaches a feature quantization model for features associated with one or more facial features descriptive of one or more faces. The quantization model includes a codebook with learned feature codes (Paragraph 7), wherein the learned feature codes can be used in place or in combination with the actual features (Paragraph 67), meaning that they are associated with one or more facial features as well. The feature codes would therefore be representing a first set of 3D geometry blocks as while Zhao is used with regards to image, the facial features themselves would be 3D geometry blocks as they are composed of 3D shapes. The codes themselves are not 3D geometry blocks, only representations of them, and as such would include features related to a 2D representation of the 3D geometry blocks.
Furthermore, the plurality of segmentations has been taught above by Le.
Zhao, Le, and the present application are analogous art because they are all in the same field of endeavor of reinforcement learning.
Zhao discloses update the first neural network using a plurality of loss terms as well as wherein at least one weight associated with the plurality of loss terms varies during the one or more training operations:
Zhao discloses that a model trainer can perform a number of techniques, including weight decays (Paragraph 84). Weight decay includes a penalty term for the loss function, wherein over subsequent training steps the weights of a plurality of terms shrink through this penalty term. These decaying weights would therefore be associated with the plurality of loss terms, such as Zhao’s realism and reconstruction losses (Paragraph 14). Through the model trainer using this loss term, the neural network is updated using the plurality loss terms with at least one weight varying during the training operations.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Le and the teachings of Zhao. This would have provided the advantage of improving accuracy and quality of reconstructed images (Zhao, Paragraph 3).
Regarding claim 2, which depends upon claim 1:
Claim 2 recites:
The computer-implemented method of claim 1, wherein each segmentation included in the plurality of segmentations specifies, for a different point in the point cloud, a different plurality of probabilities associated with a plurality of clusters of points within the first point cloud.
Le in view of Zhao disclose the method of claim 1 upon which claim 2 depends. Furthermore, regarding the limitation of claim 2:
Le in the same field of endeavor of constructive solid geometry using solid primitives teaches fitting segments of a point cloud to particular solid primitives, which are classifications of particular objects of an image for the purposes of 3D modeling (Paragraph 41). Furthermore, part of the fitting process is generated a probability of a point belonging to a fitted segment (Paragraph 41) which would be analogous to specifying for a different point in a point cloud, a different probability associated with a plurality of clusters of points within the point cloud as the cluster of points would be the segment.
Regarding claim 3, which depends upon claim 1:
Claim 3 recites:
The computer-implemented method of claim 1, wherein a first spatial feature included in the plurality of spatial features is computed by computing at least one of a mean, a weight, or a covariance associated with a cluster of points within the first point cloud
Le in view of Zhao disclose the method of claim 1 upon which claim 2 depends. However, Le does not teach the limitation of claim 3:
Zhao teaches estimating a per-location weight and determining features from the predicted weight (Paragraph 163). A per-location weight would be a weight associated with a cluster of points, and the weight is used to compute a plurality of special features.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Le and the teachings of Zhao. This would have provided the advantage of improving accuracy and quality of reconstructed images (Zhao, Paragraph 3).
Regarding claim 4, which depends upon claim 1:
Claim 4 recites:
The computer-implemented method of claim 1, wherein computing a first quantized context feature included in the plurality of quantized context features comprises: computing a first context feature based on the plurality of segmentations; computing a set of distances between the first context feature and the first set of codes; and setting the first quantized context feature equal to a first code included in the first set of codes based on the set of distances
Le in view of Zhao disclose the method of claim 1 upon which claim 4 depends. Furthermore, regarding the limitation of claim 4:
Le teaches a primitive parameter estimator that minimizes a distance metric between points in an input point cloud and its fitted primitives (Paragraph 53). The distance metric would be the first context feature based on the plurality of segmentations as it is based on the distance between the primitive and the segmentation of points it represents. The minimization of the distance metric is a computation of the distances between itself and the distance of the best-fitting primitive, and would be a computation of a set of distance between the first context feature, the distance, and the first set of codes, the final minimized distance, as it is what the distance finally is set to once minimized.
Regarding claim 5, which depends upon claim 4:
Claim 5 recites:
The computer-implemented method of claim 4, further comprising: computing a second code based on the first context feature; and replacing the first code included in the first set of codes with the second code to generate the second set of codes.
Le in view of Zhao disclose the method of claim 4 upon which claim 5 depends. However Le does not teach the limitation of claim 5:
Zhao teaches a second learned codebook which is used to process the output of a feature quantization model (Paragraph 121). The second codebook would be a second set of codes and the code used for the particular output would be the second computed code wherein the codes have been learned from features. Furthermore, the second codebook is being used in place of the first codebook, replacing the first code with the second code.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Le and the teachings of Zhao. This would have provided the advantage of improving accuracy and quality of reconstructed images (Zhao, Paragraph 3).
Regarding claim 6, which depends upon claim 1:
Claim 6 recites:
The computer-implemented method of claim 1, wherein the first neural network comprises a classification network or a segmentation network.
Le in view of Zhao disclose the method of claim 1 upon which claim 6 depends. Furthermore, regarding the limitation of claim 6:
Le teaches a primitive fitting network that uses local primitive fitting networks to evaluate local patches of points from the 3D point cloud (Paragraph 2). This would be an example of a segmentation network as the local patches would be the segments.
Regarding claim 7, which depends upon claim 1:
Claim 7 recites:
The computer-implemented method of claim 1, wherein performing one or more training operations to update the first neural network comprises: setting a plurality of variables included in a second neural network equal to the plurality of quantized context features; executing the second neural network on the plurality of spatial features and the first point cloud to evaluate a likelihood associated with reconstructing the first point cloud; and replacing a first value for a first weight included in the first neural network with a second value for the first weight that increases the likelihood associated with reconstructing the first point cloud.
Le in view of Zhao disclose the method of claim 1 upon which claim 7 depends. Furthermore, regarding the limitation of claim 7, wherein training the first neural network comprises: setting a plurality of variables included in a second neural network equal to the plurality of quantized context features:
Le teaches using global contextual information as additional input to a local primitive fitting network. The global contextual information is extracted with an encoder (Paragraph 38), wherein the encoder would be an example of quantization and then decoded such that is can used by the local primitive fitting network.
Furthermore, regarding the limitation executing the second neural network on the plurality of spatial features and the first point cloud to evaluate a likelihood associated with reconstructing the first point cloud
Le teaches using a fine detail identification network to determine a likelihood that a point is part of a small primitive (Paragraph 29). The likelihood that the point is part of a small primitive is associated with reconstructing the first point cloud as primitives are used in reconstruction.
Furthermore, the fine detail identification network may include a neural network and accepts a downsampled version of a 3D point cloud, which would comprise a plurality of spatial features and a first point cloud.
Furthermore, regarding the limitation replacing a first value for a first weight included in the first neural network with a second value for the first weight that increases the likelihood associated with reconstructing the first point cloud.
Le teaches a local primitive fitting network that uses the global contextual information generated by the global primitive fitting network to approximate a local structure of the a local patch of the 3D point cloud (Paragraph 62). That is, in order to refine the local patch’s representation a increase the likelihood associated with reconstructing the first point cloud accurately, a local primitive fitting network uses the contextual information as well as its own greater knowledge about the local patch, wherein the contextual information is the first value for the first weight and the local knowledge is the second value for the first weight.
Regarding claim 8, which depends upon claim 7:
Claim 8 recites:
The computer-implemented method of claim 7, further comprising executing one or more backpropagation operations on the first neural network to determine the second value for the first weight.
Le in view of Zhao disclose the method of claim 7 upon which claim 8 depends. However, Le does not teach the limitation of claim 8:
Zhao teaches that a loss function can be backpropagated through the models the update parameters of the model, which would include the second value for the first weight (Paragraph 83)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Le and the teachings of Zhao. This would have provided the advantage of improving accuracy and quality of reconstructed images (Zhao, Paragraph 3).
Regarding claim 9, which depends upon claim 1:
Claim 9 recites:
The computer-implemented method of claim 1, further comprising executing the trained machine learning model on a second point cloud representing a second 3D scene to generate a first representation of a first plurality of 3D geometry instances.
Le in view of Zhao disclose the method of claim 1 upon which claim 9 depends. Furthermore, regarding the limitation of claim 9:
Le teaches that following the training of the fitting network by other point clouds, the fitting system can be used take in a second point cloud representing a second 3D scene and fit primitives to it, which would be a first representation of a first plurality of 3D geometry instances (Paragraph 61).
Regarding claim 10, which depends upon claim 9:
Claim 10 recites:
The computer-implemented method of claim 9, wherein a first 3D geometry instance included in the first plurality of 3D geometry instances comprises a first instance of a first 3D geometry block included in the second set of 3D geometry blocks, and a second 3D geometry instance included in the first plurality of 3D geometry instances comprises a second instance of the first 3D geometry block or a first instance of a second 3D geometry block included in the second set of 3D geometry blocks.
Le in view of Zhao disclose the method of claim 9 upon which claim 10 depends. Furthermore, regarding the limitation of claim 10:
Le teaches fitting a 3D point cloud with 3D primitives, which would be 3D geometry instances include in a plurality of 3D geometry instances (Paragraph 61). Furthermore, these geometry instance would comprise geometry blocks include in the second set of 3D geometry blocks, which would be the types of primitives used to represent the space. Likewise, subsequent primitives, or the second 3D geometry instance include in the first plurality of 3D geometry instances would be either be the same type of primitive as thus a second instance of the first 3D geometry block or it would be a new type of primitive, and thus a first instance of a second 3D geometry block included in the second set of 3D geometry blocks.
Claims 11-13 recite a non-transitory computer readable media that parallels the method of claims 1-3 respectively. Therefore, the analysis discussed above with respect to claims 1-3 also applies to claims 11-13 respectively. Accordingly, claims 11-13 are rejected based on substantially the same rationale as set forth above with respect to claims 1-3 respectively.
Regarding claim 14, which depends upon claim 11:
Claim 14 recites:
The one or more non-transitory computer readable media of claim 11, wherein computing the plurality of quantized context features comprises: computing a plurality of context features based on the plurality of segmentations; and mapping the plurality of context features to the plurality of quantized context features based on the first set of codes.
Le in view of Zhao disclose the method of claim 11 upon which claim 14 depends. However, Le does not teach the limitation of claim 14:
Zhao teaches a feature codebook that include a feature quantization map associated with an embedding space, wherein a feature quantization model determines a location in the embedding space for the features (Paragraph 130). The features would be a plurality of context features that are mapped to a plurality of quantized context feature based on the first set of codes in the codebook.
The plurality of segmentation is taught by Le, which also teaches context features based on the plurality of segmentation. As such it would be obvious to use the context features of Le in place of the context features of Zhao in order to use Zhao’s mapping process.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Le and the teachings of Zhao. This would have provided the advantage of improving accuracy and quality of reconstructed images (Zhao, Paragraph 3).
Claims 15 recites a non-transitory computer readable media that parallels the method of claim 5. Therefore, the analysis discussed above with respect to claim 5 also applies to claim 15. Accordingly, claim 15 is rejected based on substantially the same rationale as set forth above with respect to claim 5.
Regarding claim 16, which depends upon claim 11:
Claim 16 recites:
The one or more non-transitory computer readable media of claim 11, wherein executing the first neural network on the first point cloud comprises: inputting the first point cloud into a classification network to generate a plurality of classification vectors; and normalizing each classification vector included in the plurality of classification vectors to generate the plurality of segmentations.
Le in view of Zhao disclose the method of claim 1 upon which claim 7 depends. Furthermore, regarding the limitation of claim 16 wherein executing the first neural network on the first point cloud comprises: inputting the first point cloud into a classification network to generate a plurality of classification vectors:
Le teaches generating latent vectors from the point cloud using a fitting network, which would be a form of a classification network as it fits areas of a point cloud to their best match (Paragraph 28). Furthermore, the produced object latent vector and centroid feature vector (Paragraph 28) would be classification vectors as they are produced by a classification network.
Furthermore, regarding the limitation normalizing each classification vector included in the plurality of classification vectors to generate the plurality of segmentations:
Le teaches an encoder that produces a single point vector, which would be a normalized vector (Paragraph 31). In combination with the classification vectors this would be normalizing each classification vector included in the plurality of classification vectors to generate the plurality of segmentations.
Claims 17-18 recite a non-transitory computer readable media that parallels the method of claims 7-8 respectively. Therefore, the analysis discussed above with respect to claims 7-8 also applies to claims 17-18 respectively. Accordingly, claims 17-18 are rejected based on substantially the same rationale as set forth above with respect to claims 7-8 respectively.
Regarding claim 19, which depends upon claim 11:
Claim 19 recites:
The one or more non-transitory computer readable media of claim 11, further comprising performing one or more transfer learning operations on a task-specific machine learning model that includes at least a portion of the trained machine learning model to train the task-specific machine learning model to perform a first task on one or more point clouds.
Le in view of Zhao disclose the method of claim 1 upon which claim 7 depends. Furthermore, regarding the limitation of claim 19:
Le teaches a global fitting network that utilized local fitting networks to process a portion of the point cloud (Paragraph 2). The local fitting networks’ creation would be a transfer learning operation on a task-specific machine learning model as the local fitting network process a portion of the global model for the specific area, which would be analogous to a specific task.
Claims 20 recites a system that parallels the method of claim 1. Therefore, the analysis discussed above with respect to claim 1 also applies to claim 20. Accordingly, claim 20 is rejected based on substantially the same rationale as set forth above with respect to claim 1.
Response to Arguments
Applicant’s arguments filed 21-JULY-2026 have been fully considered, but the examiner believes that not all are fully persuasive.
Regarding the applicant’s remarks on the non-final office action’s 103 rejection of the claims, the applicant argues that Le in view of Zhao do not teach the amended limitations of these claims. As such, the applicant argues that all claims dependent on the above would additionally not be obvious under 103. However, the examiner believes that Le in view of Zhao does teach the amended limitations and respectfully requests applicant’s consideration of the following:
While the examiner agrees that Le alone does not teach the amended limitation as Le does not teach multiple loss terms, let alone wherein at least one weight associated with the plurality of loss terms varies during the one or more training operations. However, the examiner believes that this limitation, as well as updating the first neural network using a plurality of loss terms, is taught by Zhao.
Zhao teaches that a model trainer can perform a number of techniques, including weight decays (Paragraph 84). Weight decay includes a penalty term for the loss function, wherein over subsequent training steps the weights of a plurality of terms shrink through this penalty term. These decaying weights would therefore be associated with the plurality of loss terms, such as Zhao’s realism and reconstruction losses (Paragraph 14). Through the model trainer using this loss term, the neural network is updated using the plurality loss terms with at least one weight varying during the training operations.
Furthermore, the arguments are silent to the teachings of the combination of Le and Zhao only addresses their teachings separately rather than integrated as the 103 rejections discuss. For this reason, the examiner believes that the arguments “Le would have to disclose the idea of performing training operations to update the neural network using multiple loss terms and based on a reconstruction of the point cloud from multiple quantized context features and the properties of geometric primitives” and “Le also would have to disclose that at least one weight associated with multiple loss terms varies during the training operations” are resolved by the integration of Zhao as described in the rejection.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALEXANDRIA JOSEPHINE MILLER whose telephone number is (703)756-5684. The examiner can normally be reached Monday-Thursday: 7:30 - 5:00 pm, every other Friday 7:30 - 4:00.
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, Mariela Reyes can be reached at (571) 270-1006. 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.
/A.J.M./Examiner, Art Unit 2142
/Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
