DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. § 119 (a)-(d). The certified copy has been filed in parent Application No. CN202311405985.1, filed on 10/26/2023.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1–6, 8–13 and 15–20 are rejected under 35 U.S.C. § 103 as being unpatentable over Guo et al. (U.S. 12,374,140 B2) in view of Lin et al. (U.S. 2023/0082597 A1).
Regarding claim 1, Guo discloses a method for training a backbone network, wherein in the backbone network, adjacent neural network layers are associated with each other and have associated weights, the method comprising:
setting a weight selection cycle, wherein the weight selection cycle comprises at least one backbone network training cycle; (Per Fig. 1 at Step S103, Guo discloses a weight of a backbone network to render a pre-training model. Guo col. 5 lines 4–7. [i]nitializing a weight of a backbone network of the vision model according to a weight of a backbone network of the trained image classification network to obtain a pre-training model.)
training the backbone network with sample data in the current weight selection cycle, (Per Fig. 2 at step S202, Guo discloses a plurality of sample images in his backbone network where image classification is extracted. Ibid. col. 6 lines 47–52. [o]btaining an unlabeled first data set, wherein the first data set includes a plurality of groups of sample images and direction information of each sample image,) and recording a cumulative weight adjustment amount (a cumulative weight adjust amount construed as a first loss and a second loss)1 for each weight in the backbone network in the current weight selection cycle, wherein the sample data comprises sample images in the field of computer vision. (Per Fig. 2 at S206, Guo adjusts the weight. Ibid. col. 10 lines 22–24. [a]djusting the weight of the backbone network of the image classification network according to the first loss and the second loss.)
However, Guo fails to specifically disclose determining, as a target weight in the backbone network, a weight for which the cumulative weight adjustment amount meets a preset condition; and adjusting the target weight in the backbone network, to complete training of the backbone network in a next weight selection cycle based on the adjusted target weight.
In related art, Lin discloses determining, as a target weight in the backbone network (a target weight in the backbone network construed as an estimation result corresponding to a sampling parameter in his target neural network), a weight for which the cumulative weight adjustment amount meets a preset condition; and (Per Fig. 9, Lin’s architecture estimator 904 determines whether the parameter—i.e., the target weight—meets a preset condition. Lin Spec. ¶202. [s]o that an estimation result of the network model corresponding to a sampling parameter obtained through next sampling is closer to or meets the preset condition.)
adjusting the target weight (adjusting the target weight construed as optimizing parameters in a network model) in the backbone network, to complete training of the backbone network in a next weight selection cycle based on the adjusted target weight. (Per Fig. 9, Lin’s parameter optimizer 902 adjusts the parameter in the loop of estimating the result such that a next sampling is closer to obtain a final target neural network. Ibid. [t]he parameter optimizer 902 may optimize the sampling mode based on the estimation result fed back by the architecture estimator 904,)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Lin into the teachings of Guo to reduce an amount of sampling data such that its efficiency is improved. Ibid. ¶11.
Regarding claim 8, Guo discloses an image processing method, comprising:
setting a weight selection cycle, wherein the weight selection cycle comprises at least one backbone network training cycle; (Per Fig. 1 at Step S103, Guo discloses a weight of a backbone network to render a pre-training model. Guo col. 5 lines 4–7. [i]nitializing a weight of a backbone network of the vision model according to a weight of a backbone network of the trained image classification network to obtain a pre-training model.)
training the backbone network with sample data in the current weight selection cycle, (Per Fig. 2 at step S202, Guo discloses a plurality of sample images in his backbone network where image classification is extracted. Ibid. col. 6 lines 47–52. [o]btaining an unlabeled first data set, wherein the first data set includes a plurality of groups of sample images and direction information of each sample image,) and recording a cumulative weight adjustment amount (a cumulative weight adjust amount construed as a first loss and a second loss) for each weight in the backbone network in the current weight selection cycle, wherein the sample data comprises sample images in the field of computer vision. (Per Fig. 2 at S206, Guo adjusts the weight. Ibid. col. 10 lines 22–24. [a]djusting the weight of the backbone network of the image classification network according to the first loss and the second loss.)
However, Guo fails to specifically disclose generating an image processing model for use in the field of computer vision by using a backbone network generated by a process for training the backbone network, wherein in the backbone network, adjacent neural network layers are associated with each other and have associated weights, the process comprising: determining, as a target weight in the backbone network, a weight for which the cumulative weight adjustment amount meets a preset condition; and adjusting the target weight in the backbone network, to complete training of the backbone network in a next weight selection cycle based on the adjusted target weight.
In related art, Lin discloses generating an image processing model for use in the field of computer vision by using a backbone network generated by a process for training the backbone network, wherein in the backbone network, adjacent neural network layers are associated with each other and have associated weights, the process comprising: (Per Fig. 3, Lin discloses a Convolutional Neural Network (CNN) where a plurality of parallel layers are embedded to extract features of image recognition. Lin Spec. ¶102. [t]he CNN may alternatively exist in a form of another network model, for example, a plurality of parallel convolutional layers/pooling layers shown in FIG. 4, and extracted features are all input to the entire neural network layer 130 for processing.)
determining, as a target weight in the backbone network (a target weight in the backbone network construed as an estimation result corresponding to a sampling parameter in his target neural network), a weight for which the cumulative weight adjustment amount meets a preset condition; and (Per Fig. 9, Lin’s architecture estimator 904 determines whether the parameter—i.e., the target weight—meets a preset condition. Lin Spec. ¶202. [s]o that an estimation result of the network model corresponding to a sampling parameter obtained through next sampling is closer to or meets the preset condition.)
adjusting the target weight (adjusting the target weight construed as optimizing parameters in a network model) in the backbone network, to complete training of the backbone network in a next weight selection cycle based on the adjusted target weight. (Per Fig. 9, Lin’s parameter optimizer 902 adjusts the parameter in the loop of estimating the result such that a next sampling is closer to obtain a final target neural network. Ibid. [t]he parameter optimizer 902 may optimize the sampling mode based on the estimation result fed back by the architecture estimator 904,)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Lin into the teachings of Guo to reduce an amount of sampling data such that its efficiency is improved. Ibid. ¶11.
Regarding claim 15, Guo discloses an electronic device, comprising:
one or more processors; and (Fig. 7, 701 a computing unit)
a storage apparatus having one or more programs stored thereon, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to train a backbone network, wherein in the backbone network, adjacent neural network layers are associated with each other and have associated weights, and the one or more programs cause the one or more processors to: (Fig. 7, 703 a RAM)
set a weight selection cycle, wherein the weight selection cycle comprises at least one backbone network training cycle; (Per Fig. 1 at Step S103, Guo discloses a weight of a backbone network to render a pre-training model. Guo col. 5 lines 4–7. [i]nitializing a weight of a backbone network of the vision model according to a weight of a backbone network of the trained image classification network to obtain a pre-training model.)
train the backbone network with sample data in the current weight selection cycle, (Per Fig. 2 at step S202, Guo discloses a plurality of sample images in his backbone network where image classification is extracted. Ibid. col. 6 lines 47–52. [o]btaining an unlabeled first data set, wherein the first data set includes a plurality of groups of sample images and direction information of each sample image,) and recording a cumulative weight adjustment amount (a cumulative weight adjust amount construed as a first loss and a second loss) for each weight in the backbone network in the current weight selection cycle, wherein the sample data comprises sample images in the field of computer vision. (Per Fig. 2 at S206, Guo adjusts the weight. Ibid. col. 10 lines 22–24. [a]djusting the weight of the backbone network of the image classification network according to the first loss and the second loss.)
However, Guo fails to specifically disclose determine, as a target weight in the backbone network, a weight for which the cumulative weight adjustment amount meets a preset condition; and adjust the target weight in the backbone network, to complete training of the backbone network in a next weight selection cycle based on the adjusted target weight.
In related art, Lin discloses determine, as a target weight in the backbone network (a target weight in the backbone network construed as an estimation result corresponding to a sampling parameter in his target neural network), a weight for which the cumulative weight adjustment amount meets a preset condition; and (Per Fig. 9, Lin’s architecture estimator 904 determines whether the parameter—i.e., the target weight—meets a preset condition. Lin Spec. ¶202. [s]o that an estimation result of the network model corresponding to a sampling parameter obtained through next sampling is closer to or meets the preset condition.)
adjust the target weight (adjusting the target weight construed as optimizing parameters in a network model) in the backbone network, to complete training of the backbone network in a next weight selection cycle based on the adjusted target weight. (Per Fig. 9, Lin’s parameter optimizer 902 adjusts the parameter in the loop of estimating the result such that a next sampling is closer to obtain a final target neural network. Ibid. [t]he parameter optimizer 902 may optimize the sampling mode based on the estimation result fed back by the architecture estimator 904,)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate the teachings of Lin into the teachings of Guo to reduce an amount of sampling data such that its efficiency is improved. Ibid. ¶11.
Regarding claim 2, Guo as modified by Lin, discloses the method, wherein the cumulative weight adjustment amount is obtained through accumulation of at least one weight adjustment amount; (Per Fig. 2 at S206, Guo adjusts the weight. Guo col. 10 lines 22–24. [a]djusting the weight of the backbone network of the image classification network according to the first loss and the second loss.) and the at least one weight adjustment amount is obtained by training the backbone network for at least one backbone network training cycle. (Per Fig. 2 at step S202, Guo discloses a plurality of sample images in his backbone network where image classification is extracted. Ibid. col. 6 lines 47–52. [o]btaining an unlabeled first data set, wherein the first data set includes a plurality of groups of sample images and direction information of each sample image,)
Regarding claim 3, Guo as modified by Lin, discloses the method, wherein the determining, as a target weight in the backbone network, a weight for which the cumulative weight adjustment amount meets a preset condition comprises:
obtaining a target number of weights with the greatest cumulative weight adjustment amount, wherein the target number is a single-cycle weight adjustment number; and (Per Fig. 9, Lin’s parameter optimizer 902 adjusts the parameter in the loop of estimating the result such that a next sampling is closer to obtain a final target neural network. Lin Spec. ¶202. [t]he parameter optimizer 902 may optimize the sampling mode based on the estimation result fed back by the architecture estimator 904,)
determining each of the target number of weights as the target weight in the backbone network. (Per Fig. 9, Lin’s architecture estimator 904 determines whether the parameter—i.e., the target weight—meets a preset condition. Ibid. [s]o that an estimation result of the network model corresponding to a sampling parameter obtained through next sampling is closer to or meets the preset condition.)
Regarding claim 4, Guo as modified by Lin, discloses the method, wherein the backbone network comprises a first structural block and a second structural block, and the first structural block is connected to the second structural block through a feature adapter; the feature adapter is configured to adjust an output of the first structural block; the feature adapter is trained synchronously during the training of the backbone network; and (Per Fig. 4, Lin discloses a neural network layer 130. Lin Spec. ¶100. [t]he CNN 100 uses the neural network layer 130 to generate an output of one required class or outputs of a group of required classes.)
the first structural block comprises a normalization layer and a multi-head attention layer, and the second structural block comprises a normalization layer and a multilayer perceptron. (Per Fig. 4, Lin discloses a pooling layer. Ibid. ¶98. The pooling layer may include an average pooling operator and/or a maximum pooling operator, to perform sampling on the input picture to obtain a picture with a relatively small size.)
Regarding claim 5, Guo as modified by Lin, discloses the method, wherein the backbone network is used to build a machine learning model in the field of computer vision; the machine learning model comprises a head network and at least one backbone network; the head network is trained synchronously during the training of the backbone network; and (Per Fig. 4, Lin discloses a neural network layer 130. Lin Spec. ¶100. [t]he CNN 100 uses the neural network layer 130 to generate an output of one required class or outputs of a group of required classes.)
the second structural block is connected to the head network through a feature adapter configured to adjust an output of the second structural block. (Per Fig. 4, Lin discloses a pooling layer. Ibid. ¶98. The pooling layer may include an average pooling operator and/or a maximum pooling operator, to perform sampling on the input picture to obtain a picture with a relatively small size.)
Regarding claim 6, Guo as modified by Lin, discloses the method, wherein the feature adapter comprises a first fully connected layer, a second fully connected layer, and a residual layer that are connected to each other; and the first fully connected layer is used to implement feature dimensionality reduction, and the second fully connected layer is used to implement feature dimensionality augmentation. (Per Fig. 4, Lin discloses a neural network layer 130. Lin Spec. ¶100. [t]he CNN 100 uses the neural network layer 130 to generate an output of one required class or outputs of a group of required classes.)
Regarding claim 9, it has been rejected in the same manner as claim 2.
Regarding claim 10, it has been rejected in the same manner as claim 3.
Regarding claim 11, it has been rejected in the same manner as claim 4.
Regarding claim 12, it has been rejected in the same manner as claim 5.
Regarding claim 13, it has been rejected in the same manner as claim 6.
Regarding claim 16, it has been rejected in the same manner as claim 2.
Regarding claim 17, it has been rejected in the same manner as claim 3.
Regarding claim 18, it has been rejected in the same manner as claim 4.
Regarding claim 19, it has been rejected in the same manner as claim 5.
Regarding claim 20, it has been rejected in the same manner as claim 6.
Allowable Subject Matter
Claims 7 and 14 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Cui et al. (U.S. 11,929,871 B2) discloses a method for generating a backbone network.
Contact
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BENEDICT LEE whose telephone number is (571)270-0390. The examiner can normally be reached 10:00-17:00 (EST).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Stephen R. Koziol can be reached at (408) 918-7630. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BENEDICT E LEE/Examiner, Art Unit 2665
/Stephen R Koziol/Supervisory Patent Examiner, Art Unit 2665
1 See his Fig. 2 at S204 and S205. In an iterative training of his image classification network, a first loss and a second loss is combined to output a final loss to determine whether it is less than a loss threshold. Ibid. col. 10 lines 29–36. [a] sum of the first loss and the second loss can be calculated as a final loss; the weight of the backbone network of the image classification network is adjusted according to the final loss, and two different loss functions can be combined to perform the self-monitoring training on the image classification network and improve the robustness and effectiveness of the image classification network.
Prosecution Insights