DETAILED ACTION
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, 5, 7-12 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Zhou et al (US 10970518), in view of Bello et al (US 20210350179), in further view of Blondel et al (US 20200242330).
Regarding claim 1, Zhou discloses a method for learning representations from clouds of points data (col 1 lines 50-54 methods, systems and/or techniques for implementing a 3D voxel feature learning/detection network that may be configured to learn voxel features automatically from raw point cloud data (e.g., a LiDAR point cloud)), comprising:
encoding clouds of points data into at least one representation by creating at least one tensor representation out of the clouds of points data (col 6 lines 47-67 divide a point cloud into equally spaced 3D voxels, encode each voxel via stacked VFE layers to determine a sparse 4D tensor representation of the point cloud).
Zhou fails to teach where Bello teaches using a loss function that utilizes a noisy reconstruction for reducing overfitting (¶142 The 4Dsurvival network 2 was based on a denoising autoencoder (DAE), which is a type of autoencoder which aims to extract more robust latent representations 12 by corrupting the input, for example vector x representing a time resolved three-dimensional model 4 with stochastic noise; A loss function 16 was used in the form of a hybrid loss function having a contribution from a reconstruction loss 19 and a contribution from a prediction loss 20; To safeguard against overfitting on the training set 5, dropout and L.sub.1 regularization were used in order to yield a robust prediction model).
Zhou fails to further teach where Blondel teaches comparing the at least one representation with other representations of clouds of points data stored in a database to determine a match (¶168 the reconstructed point cloud 5050 and the database point cloud 6050 will fit in all of the 3 features position/rotation/scale and the calculation of the distance between the points clouds can be computed. Search for the minimal distances will give the best match in the database as shown on FIG. 19E).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of using a loss function that utilizes a noisy reconstruction for reducing overfitting from Bello, and the teaching of comparing the at least one representation with other representations of clouds of points data stored in a database to determine a match from Blondel into the method for learning representations as disclosed by Zhou. The motivation for doing this is to improve training a machine learning model and further to improve object recognition.
Regarding claim 5, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, wherein the encoding is performed together with at least one data augmentation technique, wherein the at least one data augmentation technique includes at least one rotation, translation, and/or other type of distortion (Zhou col 16 lines 63-65 the system may employ global rotation to all ground truth boxes b.sub.i and to the whole point cloud M in some embodiments).
Regarding claim 7, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, wherein the loss function cooperates with a machine learning model (Bello abstract training a machine learning model; ¶142 The prediction loss 20 for training the exemplary machine learning model 2 was inspired by the Cox proportional hazards model; see Fig. 1 e.g. machine learning model 2 cooperating with loss function 16). The motivation to combine the references is discussed above in the rejection for claim 1.
Regarding claim 8, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, wherein the loss function cooperates with a Neural Network (NN) or Artificial Neural Network (ANN) (Bello ¶84 the trained machine learning model 2 (FIG. 2) takes the form of a neural network; ¶142 A loss function 16 was used in the form of a hybrid loss function having a contribution from a reconstruction loss 19 and a contribution from a prediction loss 20; see Fig. 1 e.g. machine learning model 2 cooperating with loss function 16). The motivation to combine the references is discussed above in the rejection for claim 1.
Regarding claim 9, the combination of Zhou, Bello, and Blondel disclose the method according to claim 8, wherein the clouds of points data are passed through the NN or ANN (Zhou col 3 lines 10-17 in order to determine a voxel feature for a respective voxel of a point cloud, the fully connected neural network comprising the voxel feature encoding layers may determine one or more point-wise features from a set of points included in a portion of the point cloud corresponding to a voxel).
Regarding claim 10, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, wherein the noisy reconstruction is produced by a decoder or encoder with dropout sampling (Bello ¶142 To safeguard against overfitting on the training set 5, dropout and L.sub.1 regularization were used in order to yield a robust prediction model.). The motivation to combine the references is discussed above in the rejection for claim 1.
Regarding claim 11, the combination of Zhou, Bello, and Blondel disclose the method according to claim 10, wherein the decoder or encoder produces noisy versions of the clouds of points data (Bello ¶95 When the machine learning model 2 includes a de-noising autoencoder, the input layer 9 and/or one or more encoding hidden layers 13 may implement a mask configured to apply stochastic noise to the inputs), wherein the noisy versions are passed back to the decoder or encoder for generating embeddings of the inputs (Bello ¶95 the machine learning model 2 may include hidden layers 13, 16 and an encoding layer 11 which form a de-noising autoencoder; ¶92 & ¶187 machine learning model may include an encoding layer 11 which encodes a latent representation 12 of cardiac motion). The motivation to combine the references is discussed above in the rejection for claim 1.
Regarding claim 12, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, wherein the clouds of points data comprise 2D clouds of points (Zhou col 5 lines 32-33 3D point cloud data is condensed into 2D images that can be analyzed to determine objects in the point cloud).
Regarding claim(s) 15 (drawn to a system):
The rejection/proposed combination of Zhou, Bello, and Blondel, explained in the rejection of method claim(s) 1, anticipates/renders obvious the steps of the system of claim(s) 15 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1 is/are equally applicable to claim(s) 15.
Claim(s) 2-4 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Zhou, Bello, and Blondel as applied to claim 1 above, and further in view of Xu et al (NPL: SqueezeSegV3: Spatially-Adaptive Convolution for Efficient Point-Cloud Segmentation).
Regarding claim 2, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, but fail to teach where Xu teaches wherein the clouds of points data or points of the clouds of points data are projected into tensors having a predefined height, width, and channel depth (pg. 5-6 3 Spherical Projection of LiDAR Point-Cloud: To process a LiDAR point-cloud efficiently, Wu et al. [56] proposed a pipeline (shown in Figure 1) to project a sparse 3D point cloud to a 2D LiDAR image; This way, we can represent a LiDAR point cloud as a LiDAR image with the shape of (h, w, 5)).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the clouds of points data or points of the clouds of points data are projected into tensors having a predefined height, width, and channel depth from Xu into the method as disclosed by the combination of Zhou, Bello, and Blondel. The motivation for doing this is to improve 2D/3D object detection, multi-modal fusion, simultaneous localization and mapping, and point-cloud segmentation.
Regarding claim 3, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, but fail to teach where Xu teaches wherein the clouds of points data or points of the clouds of points data are converted into 2D locations that lie in a defined height and width (pg. 5-6 3 Spherical Projection of LiDAR Point-Cloud: To process a LiDAR point-cloud efficiently, Wu et al. [56] proposed a pipeline (shown in Figure 1) to project a sparse 3D point cloud to a 2D LiDAR image; (h, w) are the height and width of the desired projected 2D map).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the clouds of points data or points of the clouds of points data are converted into 2D locations that lie in a defined height and width from Xu into the method as disclosed by the combination of Zhou, Bello, and Blondel. The motivation for doing this is to improve 2D/3D object detection, multi-modal fusion, simultaneous localization and mapping, and point-cloud segmentation.
Regarding claim 4, the combination of Zhou, Bello, Blondel, and Xu disclose the method according to claim 3, wherein associated features of several points of the clouds of points data or each point of the clouds of points data are encoded into a c dimension directed along a length of a channel (Xu pg. 5-6 3 Spherical Projection of LiDAR Point-Cloud: For each point projected to (p, q), we use its measurement of (x, y, z, r) and intensity as features and stack them along the channel dimension.). The motivation to combine the references is discussed above in the rejection for claim 3.
Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Zhou, Bello, and Blondel as applied to claim 1 above, and further in view of Ren et al (US 20210124985).
Regarding claim 6, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, but fails to teach where Ren teaches wherein the loss function is an unsupervised loss function (¶44 an unsupervised loss function).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the loss function is an unsupervised loss function from Ren into the method as disclosed by the combination of Zhou, Bello, and Blondel. The motivation for doing this is to improve computer vision applications.
Claim(s) 13-14 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Zhou, Bello, and Blondel as applied to claim 1 above, and further in view of Guo et al (US 20180121713).
Regarding claim 13, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, but fails to teach where Guo teaches wherein the method is used for matching of clouds of points for biometric matching, fingerprint matching or face matching, wherein a cloud of points represents a biometric sample, a fingerprint or a face (¶56 For example, the second alignment may include performing point cloud matching between the partial face depth map and the full face data).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the method is used for matching of clouds of points for biometric matching, fingerprint matching or face matching, wherein a cloud of points represents a biometric sample, a fingerprint or a face from Guo into the method as disclosed by the combination of Zhou, Bello, and Blondel. The motivation for doing this is to improve systems and methods for verifying a face.
Regarding claim 14, the combination of Zhou, Bello, Blondel, and Guo disclose the method according to claim 13, wherein the matching is obtained by finding a most similar cloud of points or clouds of points of the clouds of points data (Guo ¶56 the second aligner 114 may perform a second alignment of the partial face depth map and the full face data based on the first alignment (e.g., the alignment, the transformation, coarse alignment, the aligned partial face depth map, etc.). For example, the second alignment may include performing point cloud matching between the partial face depth map and the full face data; ¶57 For example, each subset (e.g., small subset) of neighboring data points of the partial depth map may be aligned to corresponding points in the full depth map in an iterative procedure; This procedure may iterate until an objective function is minimized. This second (e.g., local, fine, etc.) alignment may ensure improved (e.g., the best) registration between the partial face depth map and the full face depth map.). The motivation to combine the references is discussed above in the rejection for claim 13.
Claim(s) 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Zhou, Bello, and Blondel as applied to claim 1 above, and further in view of Mao et al (US 20200202145).
Regarding claim 16, the combination of Zhou, Bello, and Blondel disclose the method according to claim 1, but fail to teach where Mao teaches wherein the encoding the clouds of points data comprises: projecting minutiae from the clouds of points data into an image matrix (¶62-63 can generate projections of point cloud data; system can represent each projection as a matrix of data, with each element of the matrix corresponding to a location on the projection plane); and processing, using an encoder, the image matrix to obtain an embedding (¶56 the channel encoders 310a-n each process a different one of the first neural network inputs for a sensor channel corresponding to the encoder; object classifier neural network 302 is configured to process the first neural network inputs for patches 316a-n of the object of interest and a corresponding feature vector 322 from the context map to generate an object classification 324; ¶68 At stage 610, a context embedding neural network (e.g., network 308) processes the wide-view representation of the environment to generate a context map. The context map includes a collection of feature vectors, each corresponding to a different region of the environment encompassed by the wide-view representation).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the encoding the clouds of points data comprises: projecting minutiae from the clouds of points data into an image matrix and processing, using an encoder, the image matrix to obtain an embedding from Mao into the method as disclosed by the combination of Zhou, Bello, and Blondel. The motivation for doing this is to improve classifications of objects represented in data.
Regarding claim 17, the combination of Zhou, Bello, Blondel and Mao disclose the method according to claim 16, wherein the comparing of the at least one representation with the another representation of clouds of points data to determine a match comprises: computing a distance metric between the at least one representation and the other representations of clouds of point data stored in the database, wherein computing the distance metric comprises computing a respective score for each representation of clouds of points data stored in the database (Blondel ¶168 Thus, the reconstructed point cloud 5050 and the database point cloud 6050 will fit in all of the 3 features position/rotation/scale and the calculation of the distance between the points clouds can be computed. Search for the minimal distances will give the best match in the database as shown on FIG. 19E.). The motivation to combine the references is discussed above in the rejection for claim 1.
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Zhou, Bello, Blondel, and Mao as applied to claim 17 above, and further in view of Braham et al (US 20150095079).
Regarding claim 18, the combination of Zhou, Bello, Blondel and Mao disclose the method according to claim 17 including representations of clouds of points data stored in the database (Blondel ¶168 the reconstructed point cloud 5050 and the database point cloud 6050 will fit in all of the 3 features position/rotation/scale and the calculation of the distance between the points clouds can be computed), but fail to teach where Braham teaches ranking each of the other representations of clouds of points data stored in the database based on the respective score to generate a ranked list of representations (¶25 an exact match ranks the highest an exact match of the keywords and input skills of the symptom skills table 131 ranks the highest of available skills to address the problem report files 130); and applying an exact matching algorithm on the ranked list of representations to determine the match with the at least one representation (¶25 an exact match ranks the highest an exact match of the keywords and input skills of the symptom skills table 131 ranks the highest of available skills to address the problem report files 130).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of ranking each of the other representations of clouds of points data stored in the database based on the respective score to generate a ranked list of representations and applying an exact matching algorithm on the ranked list of representations to determine the match with the at least one representation from Braham into the method as disclosed by the combination of Zhou, Bello, Blondel and Mao. The motivation for doing this is to improve ranking determined by closeness of match between data.
Response to Arguments
Applicant’s arguments with respect to the rejection under 35 U.S.C. 101 have been fully considered and are persuasive. The rejection under 35 U.S.C. 101 of claims 1-15 has been withdrawn.
Applicant’s arguments with respect to claim(s) 1-15 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.
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 KEVIN KY whose telephone number is (571)272-7648. The examiner can normally be reached Monday-Friday 9-5PM.
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 at 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.
/KEVIN KY/Primary Examiner, Art Unit 2671
Prosecution Insights