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
Receipt is acknowledged of certified copies of papers submitted under 35 U.S.C. 119(a)-(d), which papers have been placed of record in the file.
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 submissions, filed on march 04TH, 2026, have been entered.
Status of Claims
Claims 1, 4, 12-17, 20-23 and 26-33 are pending, claims 1, 4, 16-17, 20-21, 23 and 26-27 have been amended, claims 2-3, 5-11, 18-19, 24-25 have been canceled, claims 28-33 have been added. Claims 1, 4, 12-17, 20-23 and 26-33 remains rejected.
Response to Argument(s)
In view of the Amendments to independent claims 1, 17 and 23, the previously applied prior art rejections are withdrawn. Applicants’ arguments are rendered moot in view of the new grounds of rejection set forth below.
Regarding the Applicants’ argument to the 101 rejections:
In pages 10-13 of the remarks, the Applicants argue that the disclosure in paragraphs 4, 8, 49, 58 and 145 solve the problem that different feature analysis network models require different types of image features, which a single shared feature extraction model cannot satisfy for all tasks, therefore, by using identification information to route features to appropriate analysis models, embodiments of the present disclosure architecture also increases flexibility and scalability for network updates, allowing models to be updated on select devices rather than all devices, while avoiding resource waste. Therefore, claims 1, 4, 12-17 and 20-23 (previously rejected under 101) are directed to particular improvement in the capabilities of computing device, and the claims are not directed to abstract idea.
Moreover, claims 1, 4, 12-17 an 20-23 provide limitations directed to a practical application. For instance claims 1, 4, 12-17, and 202-23 provide an improvement in performance of a computing device in multi-task image processing using identification information to route features from multiple feature extraction models to appropriate analysis models.
Examiner’s reply:
The examiner respectfully disagrees with the Applicants’ arguments and find them to be incommensurate with the scope of the claims. Importantly, the Applicants are respectfully reminded that the claims are construed based on BRI (broadest reasonable interpretation) in light of the specification, meaning that the specification is used to assist understanding of the claims, however, cannot be imported its teachings to be the instant scope of the claims. Therefore, by stating that the disclosure’s paragraphs 4, 8, 49, 58 and 145 teaching a solving of a problem, such as stated in the Applicants’ remarks, then have the claims 1, 4, 12-17 and 20-23 carry these ideas without having these particular ideas reflected in the claims. Particularly, nowhere in the claims recite or closely reflect “different feature analysis network models require different types of image features, which a single shared feature extraction model cannot satisfy for all tasks, therefore, by using identification information to route features to appropriate analysis models, embodiments of the present disclosure architecture also increases flexibility and scalability for network updates, allowing models to be updated on select devices rather than all devices, while avoiding resource waste.”
Furthermore, the examiner finds the claims to recite abstract ideas, such as laid out in the 101 analysis of the previous Office Actions. Importantly noted that: under MPEP 2106.04(a)(2)(III), mental process (thinking) “can be performed in the human mind, or by a human using a pen and paper" to be an abstract idea. CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1372, 99 USPQ2d 1690, 1695 (Fed. Cir. 2011): "methods which can be performed mentally, or which are the equivalent of human mental work, are unpatentable abstract ideas the ‘basic tools of scientific and technological work’ that are open to all." (citing Gottschalk v. Benson, 409 U.S. 63, 175 USPQ 673 [1972]). See also Mayo Collaborative Servs. v. Prometheus Labs. Inc., 566 U.S. 66, 71, 101 USPQ2d 1961, 1965 ("mental processes and abstract intellectual concepts are not patentable, as they are the basic tools of scientific and technological work’" (quoting Benson, 409 U.S. at 67, 175 USPQ at 675).
The courts do not distinguish between mental processes that are performed entirely in the human mind and mental processes that require a human to use a physical aid (e.g., pen and paper or a slide rule) to perform the claim limitation. See, e.g., Benson, 409 U.S. at 67, 65, 175 USPQ at 674-75, 674; Synopsys, Inc. v. Mentor Graphics Corp., 839 F.3d 1138, 1139, 120 USPQ2d 1473, 1474 (Fed. Cir. 2016).
Nor do the courts distinguish between claims that recite mental processes performed by humans and claims that recite mental processes performed on a computer, generic circuit or device, or the likes. See " Versata Dev. Group v. SAP Am., Inc., 793 F.3d 1306, 1335, 115 USPQ2d 1681, 1702 (Fed. Cir. 2015). See also Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1318, 120 USPQ2d 1353, 1360 (Fed. Cir. 2016) (‘with the exception of generic computer-implemented steps, there is nothing in the claims themselves that foreclose them from being performed by a human, mentally or with pen and paper.’’).
Because both product/device and process claims may recite a "mental process", the phrase "mental processes" should be understood as referring to the type of abstract idea, and not to the statutory category of the claim. The courts have identified numerous product claims as reciting mental process-type abstract ideas, for instance the product claims to computer systems and computer-readable media in Versata Dev. Group. v. SAP Am., Inc., 793 F.3d 1306, 115 USPQ2d 1681 (Fed. Cir. 2015). Therefore, these are merely steps, considered to be mental process abstract ideas of the claims limitations, such as analyzed below under the 101 rejection section.
Regarding the second central point of the Applicants’ 101 arguments, the examiner finds the claims not to reflect a practical application of improving capabilities of a computing device. Importantly, there is no alteration into the functionalities, or structure of a computer or computing device, the claims still use generic conventional well-known computer with well-known components performing well-known functions such as a processor executing the instructions of the invention (storable programmable instructions, the steps of the claims) stored in a memory of a computer. Therefore, there is no improvement in a computing device as stated as one of the requirements for step 2A prong 2 of the 101 requirements.
PNG
media_image1.png
524
482
media_image1.png
Greyscale
Again, similarly, the Applicant’s arguments for this point are not commensurate with the scope of the claims, since the claims are construed based on BRI in light of the specification as the first requirement of the 101 analysis.’
Therefore, the 101 rejections remain, and applied for the added claims as well. See 101 rejection section below for more details.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1, 4, 12-17, 20-23 and 26-33 are rejected under 35 U.S.C. 101
Regarding independent claim 1 and its dependent claims 4 and 12-16 and 28-33,
Step 1 Analysis: Claim 1 is directed to a method/process, which falls within one of the four statutory categories.
Step 2A Prong 1 Analysis: Claim 1 recites, in part, “identifying feature information to obtain identification information of the feature information; select a feature analysis network model corresponding to an identification information to process the feature information.” These limitations as drafted, are processes that, under broadest reasonable interpretation, covers the performance of the limitation in the mind which falls within the “Mental Processes” grouping of abstract ideas. The limitations of:
“identifying feature information to obtain identification information of the feature information” are steps, under BRI (broadest reasonable interpretation) scope, understood to be mental processes wherein the human mind can observe an image and evaluate its information to identify feature information to obtain identification information and select corresponding feature analysis network model;
“select a feature analysis network model corresponding to an identification information to process the feature information” which is a mental process abstract idea wherein, the human mind can select a model to use based on some observable desirable condition, criteria such as recited in the claim.
Accordingly, the claim recites an abstract idea.
Step 2A Prong 2 Analysis: This judicial exception is not integrated into a practical application. particular, the claim recites the following additional element(s) –
storing, to a memory of a first device, a plurality of feature extraction network models, wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image;
wherein the identification information of the feature information comprises a first identifier of the least one of the plurality of feature extraction network models
processing, by one or more processors of the first device and using at least one of the plurality of feature extraction network models, the to-be-processed image to extract feature information of the to-be-processed image;
sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model of the neural network;
The additional elements are indicated, are steps considering data gathering insignificant extra-solution activities of extracting data/information of a model and an image, and sending these data/information to another device from a first device (generic high level of generality recitation of a device) and generic high level of generality of neural network without further limiting, in details, how the neural network structured and function to arrive at such outcomes. Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claim as a whole is directed to an abstract idea. Please see MPEP §2106.04.(d).III.C.
The additional elements include steps of insignificant extra-solution/post-solution activities of data gathering, data generating, data transmitting, etc. [acquiring data/information, transmitting data/info., outputting data/information, displaying data/info., converting data/info., generating data/info., etc.] .
The additional element includes generic neural network/machine learning model(s)/classifier recited at high level of generality without limiting further, in details, on how the neural network/machine learning model(s) function to arrive at such output, these additional elements are recited as a mere attempt to implement the abstract ideas/judicial exceptions using generic neural network/machine learning models.
Step 2B Analysis: there are no additional elements that amount to significantly more than the judicial exception. Please see MPEP §2106.05. The claim is directed to an abstract idea.
For all of the foregoing reasons, claim 1 does not comply with the requirements of 35 USC 101.
Accordingly, the dependent claims 4 and 12-16, 28-33 do not provide elements that overcome the deficiencies of the independent claim 1. Moreover, claims 4 and 12-16, 28-33 recites further additional elements, under Step 2A Prong 2, to be insignificant extra-solution activities for the “obtaining….” steps are merely steps involving obtaining data/information regarding an image, feature information or a model, by BRI, are data gathering steps; “wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network” indicates a selection, therefore, only one option is the instant scope of the claim, moreover, any of which is just generic well known neural network and component, hence, not an indication of an integration of the judicial exceptions into a practical application nor considered significantly more; and the step of “using….” are merely steps that, under Step 2A Prong 2, to be using information/data of insignificant post-solution activities, these are not an indication of an integration of the judicial exceptions into a practical application; moreover, the limitation of “wherein…comprises…” are just limitation that provide general further specification to the abstract ideas, additional elements that they each depend on hence, still abstract ideas, additional elements; claims 28-33 recite additional elements of generic neural network and neural network components well-known in the art recited at high level of generality without further limiting, in details, how the neural network work to arrive at such output, the claims also recites additional elements of insignificant extra-solution activities of data gathering, data inputting, data outputting, and generic devices and device components such as a charge coupled device; claim 14 recites further a limitation of a “determining….” which is a mental process abstract idea wherein the human mind can evaluate the information/data to determine the feature analysis network model for processing of the information/data.
Accordingly, the dependent claims 4 and 12-16 and 28-33 are not patent eligible under 101.
Regarding independent claim 17 and its dependent claim 20-22,
The independent claim 17 recites analogous limitations to the independent claim 1 hence, can be analyzed under the same approach as for claim 1 above to be 101 ineligible, moreover, claim 17 recites further additional elements, under Step 2A Prong 2 and Step 2B, of an electronic device, at least one memory configured to store instructions, at least one processor coupled to the at least one memory and configured to execute the instructions to cause the electronic device to perform the steps; however, these additional elements are generic computer and its components to perform generic well-known functions in the art hence, not indicative of an integration of the judicial exceptions into a practical application nor being considered significantly more.
Regarding the dependent claims 20-22, these claims carry analogous limitations to the dependent claims 12-16 and 28-33 which can be analyzed under the same approach to be 101 ineligible; these claims recite limitations of “obtain….,” “use….” which were merely obtaining and using of information/data to be insignificantly extra-solution activity additional elements, and the “wherein….comprises…” limitations are limitations of general further specification to the additional elements and abstract ideas hence, still are just insignificant additional elements and abstract ideas.
Regarding independent claim 23 and its dependent claim 26-27,
The independent claim 23 recites analogous limitations to the independent claim 1 hence, can be analyzed under the same approach as for claim 1 above to be 101 ineligible, moreover, claim 23 recites further additional elements, under Step 2A Prong 2 and Step 2B, of a computer program product comprising instructions that are stored on a computer-readable medium executed by a processor, cause a first electronic device to; however, these additional elements are generic computer and its components to perform generic well-known functions in the art hence, not indicative of an integration of the judicial exceptions into a practical application nor being considered significantly more.
Regarding the dependent claims 26-27, these claims carry analogous limitations to the dependent claims 12-16 and 28-33 which can be analyzed under the same approach to be 101 ineligible; these claims recite limitations of “obtain….,” “use….” which were merely obtaining and using of information/data to be insignificantly extra-solution activity additional elements, and the “wherein….comprises…” limitations are limitations of general further specification to the additional elements and abstract ideas hence, still are just insignificant additional elements and abstract ideas.
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.
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, 12, 14-15, 17, 21, 23, 27-28, 30-31 and 33 are rejected under 35 U.S.C. 103 as being unpatentable over Matthias Feurer et. al. (“Initializing Bayesian Hyperparameter Optimization via Meta-Learning, 2015, Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 29, No. 1 [2015]: Twenty-Nonth AAAI Conference on Artificial Intelligence, AAAI Technical Track: Heuristic Search and Optimization” hereinafter as “Feurer”) in view of Joaquin Vanschoren et. al. (“Meta-Learning and Algorithm Selection Workshop at ECAL, 2014, 21st European Conference on Artificial Intelligence” hereinafter as “Vanschoren”) and Rastislav Lukac et. al. (“US 2019/0325580 A1” hereinafter as “Lukac”).
Regarding claim 1, Fuerer discloses an image processing method (abstract), comprising: storing, to a memory of a first device, a plurality of feature extraction network models (abstract discloses the invention is based on computing of a device, hence, include the use of a computer indicating storing of the data the information all being used in the invention in a memory of a computer device, including the feature extraction network models; page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model]); processing, by one or more processors of the first device and using at least one of the plurality of feature extraction network models, the to-be-processed image to extract, feature information of the to-be-processed image (page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model] sine the claim simply recites “feature extraction network model” which doesn’t explicitly indicates a machine learning or a network of a neural network or not, therefore, by BRI, the examiner reads it to fall within the scope of Fuerer’s algorithm 2 wherein the metafeatures are calculated from the datasets [feature extraction] based on an algorithm with specific steps [network model]; the processor performing the function of algorithm 2 can be understood to be the first device as claimed; furthermore, page 1131 and FIG. 1 shows that the dataset includes dataset from OpenML and used for classification task such as liver disorders classification which can be understood to include dataset of images [any can be understood to the recited “to-be-processed image”]); identifying, by the first device, the extracted feature information to obtain identification information of the feature information (page 1130, 2nd col., 1st par., discloses a metric is calculated for the dataset based on the metafeatures, therefore, the measure information used to obtain the metric is analogous to the identification information as claimed, by BRI), wherein the identification information of the feature information comprises a first identifier of the at least one of the plurality of feature extraction network models (the processing of metafeature computation and algorithm 2 to be the pre-stored feature extraction network model; moreover, the metric’s measures obtained from the algorithm 2 is the identification information discussed above in claim 1 as disclosed in page 1130, 2nd col.; therefore, the metric itself can be understood as the identifier as claimed by BR); and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information (page 1129 discloses algorithm 1 to select the model optimal for performing the function most suitable for the dataset based on the metric’s measure information obtained and the metafeature which is analogous to the claimed limitation, the processor performing algorithm 1 can be understood to be the second device as claimed, by BRI).
However, Fuerer does not explicitly disclose at least one pre-stored feature extraction network model of a neural network; a feature analysis network model of the neural network; wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image.
In the same field of Meta Learning and Algorithm (title, Vanschoren) Vanschoren discloses at least one pre-stored feature extraction network model of a neural network; a feature analysis network model of the neural network (algorithm 2, and the selected models as disclosed in section 3, 1st 4 paragraphs, are analogous to the algorithm 2 and the selected model of Fuerer, moreover, Vanschoren further teaches that the algorithm 2 and the processing of the selection is for neural network task such as disclosed in section 1, 1st par. hence, is analogous to “of a neural network” as claimed, as further explained in section 2 of page 25).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer to perform extracting, by a first device and using at least one pre-stored feature extraction network model, feature information of a to-be-processed image; identifying, by the first device, the feature information to obtain identification information of the feature information; and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information; wherein the at least one pre-stored feature extraction network model is of a neural network and the feature analysis network model if of the neural network as taught by Vanschoren to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to apply machine learning to novel dataset in an improve and efficient way (abstract, Vanschoren).
However, Fuerer in view of Vanschoren does not explicitly disclose wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image.
In the same field of dataset similarity measurement (abstract and [0039], Lukac) Lukac discloses wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model ([0076] discloses a refinement process to use one or more of feature extraction, matching, corner detection, template matching, semantic segmentation [semantic segmentation network], object detection [target detection network], image registration, trained classifier [image classification network] to adaptively maximize similarities in the overlapping region in images from two neighboring cameras [which is used to maximize similarities between two datasets which is analogous to Fuerer’s page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model] sine the claim simply recites “feature extraction network model”], which Fuerer uses the algorithm to configure optimalization for similar dataset [page 1130, 1st par.]); receiving, from a sensor of the first device, a to-be-processed image ([0075] discloses the image are obtained from a camera of a device).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren to perform extracting, by a first device and using at least one pre-stored feature extraction network model, feature information of a to-be-processed image; identifying, by the first device, the feature information to obtain identification information of the feature information; and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information; wherein the at least one pre-stored feature extraction network model is of a neural network and the feature analysis network model if of the neural network; wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model;receiving, from a sensor of the first device, a to-be-processed image as taught by Lukac to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to uses different models to perform projection refinement to enhance similarities between datasets for detection to perform alignment processing more efficiently ([0075-0077], Lukac).
Regarding claim 12, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 1, wherein identifying the feature information comprises (as discussed above in claim 1): obtaining the identification information according to an image processing task (page 1130, 2nd column, as discussed above of the obtaining of the metafeatures and the metric information, which is according to the classification task to optimize relevant machine learning framework [page 1131, 1st col, “ML Algorithm and Hyperparameters” section, 1st par.], by BRI, is analogous to the claimed limitation).
Regarding claim 14, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 1, further comprising: obtaining, by the second device, the feature information and the identification information from the first device (as discussed above in claim 1, the algorithm 1 uses the information output from the algorithm 2, therefore, it can be understood as the second device obtain the feature information and the identification for processing, by BRI); determining, by the second device based on the identification information, the feature analysis network model for processing the feature information (as discussed above in claim 1, the algorithm 2 is to select the optimal machine learning for the dataset which is analogous to the feature analysis network model as claimed, by BRI); and inputting, by the second device, the feature information to the feature analysis network model to obtain an image processing result (and algorithm 1 processes the feature information to be input into the selected machine learning model to obtain the image processing result, such as discussed above in claim 1 and disclosed in page 1130)..
Regarding claim 15, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 14, wherein determining the feature analysis network model comprises (as discussed above in claims 1 and 14): obtaining a correspondence between the identification information and the feature analysis network model (page 1130, 1st col., 2nd par., discloses based on the metric a distance between the datasets is obtained, therefore, the distance here is analogous to “correspondence” as claimed, since the increasing distance also correspond to the sorting of the machine learning configurations optimal for use on the dataset [between the identification information and the feature analysis network model as claimed, since the metric is analogous to the identification information and the machine learning configuration here is analogous to the feature analysis network model]); and using, based on the correspondence, the feature analysis network model corresponding to the identification information as the feature analysis network model for processing the feature information (as based on the distance and the sorting, the optimal machine learning mode configuration is used for the dataset, by BRI, is analogous to the claimed limitation).
Regarding claim 17, Fuerer discloses a first electronic device, comprising at least one memory configured to store instructions; and one or more processors coupled to the at least one memory and configured to execute the instructions to cause the first electronic device to (abstract discloses the processing is for machine learning which can be understood to include the use of a computer which includes a memory storing instructions to be executed by a processor): store, to the at least one memory, a plurality of feature extraction network models (abstract discloses the invention is based on computing of a device, hence, include the use of a computer indicating storing of the data the information all being used in the invention in a memory of a computer device, including the feature extraction network models; page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model]); process, by the one or more processors and using at least one of the plurality of feature extraction network models, the to-be-processed image to extract, feature information of the to-be-processed image (page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model] sine the claim simply recites “feature extraction network model” which doesn’t explicitly indicates a machine learning or a network of a neural network or not, therefore, by BRI, the examiner reads it to fall within the scope of Fuerer’s algorithm 2 wherein the metafeatures are calculated from the datasets [feature extraction] based on an algorithm with specific steps [network model]; the processor performing the function of algorithm 2 can be understood to be the first device as claimed; furthermore, page 1131 and FIG. 1 shows that the dataset includes dataset from OpenML and used for classification task such as liver disorders classification which can be understood to include dataset of images [any can be understood to the recited “to-be-processed image”]); identify, the feature information, to obtain identification information of the feature information (page 1130, 2nd col., 1st par., discloses a metric is calculated for the dataset based on the metafeatures, therefore, the measure information used to obtain the metric is analogous to the identification information as claimed, by BRI), wherein the identification information of the feature information comprises a first identifier of the at least one of the plurality of feature extraction network models (the processing of metafeature computation and algorithm 2 to be the pre-stored feature extraction network model; moreover, the metric’s measures obtained from the algorithm 2 is the identification information discussed above in claim 1 as disclosed in page 1130, 2nd col.; therefore, the metric itself can be understood as the identifier as claimed by BR); and send the feature information and the identification information to a second device to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information (page 1129 discloses algorithm 1to select the model optimal for performing the function most suitable for the dataset based on the metric’s measure information obtained and the metafeature which is analogous to the claimed limitation, the processor performing algorithm 1 can be understood to be the second device as claimed, by BRI).
However, Fuerer does not explicitly disclose at least one pre-stored feature extraction network model of a neural network; a feature analysis network model of the neural network; wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image.
In the same field of Meta Learning and Algorithm (title, Vanschoren) Vanschoren discloses at least one pre-stored feature extraction network model of a neural network; a feature analysis network model of the neural network (algorithm 2, and the selected models as disclosed in section 3, 1st 4 paragraphs, are analogous to the algorithm 2 and the selected model of Fuerer, moreover, Vanschoren further teaches that the algorithm 2 and the processing of the selection is for neural network task such as disclosed in section 1, 1st par. hence, is analogous to “of a neural network” as claimed, as further explained in section 2 of page 25).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer to perform extracting, by a first device and using at least one pre-stored feature extraction network model, feature information of a to-be-processed image; identifying, by the first device, the feature information to obtain identification information of the feature information; and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information; wherein the at least one pre-stored feature extraction network model is of a neural network and the feature analysis network model if of the neural network as taught by Vanschoren to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to apply machine learning to novel dataset in an improve and efficient way (abstract, Vanschoren).
However, Fuerer in view of Vanschoren does not explicitly disclose wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image.
In the same field of dataset similarity measurement (abstract and [0039], Lukac) Lukac discloses wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model ([0076] discloses a refinement process to use one or more of feature extraction, matching, corner detection, template matching, semantic segmentation [semantic segmentation network], object detection [target detection network], image registration, trained classifier [image classification network] to adaptively maximize similarities in the overlapping region in images from two neighboring cameras [which is used to maximize similarities between two datasets which is analogous to Fuerer’s page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model] sine the claim simply recites “feature extraction network model”], which Fuerer uses the algorithm to configure optimalization for similar dataset [page 1130, 1st par.]); receiving, from a sensor of the first device, a to-be-processed image ([0075] discloses the image are obtained from a camera of a device).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren to perform extracting, by a first device and using at least one pre-stored feature extraction network model, feature information of a to-be-processed image; identifying, by the first device, the feature information to obtain identification information of the feature information; and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information; wherein the at least one pre-stored feature extraction network model is of a neural network and the feature analysis network model if of the neural network; wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model;receiving, from a sensor of the first device, a to-be-processed image as taught by Lukac to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to uses different models to perform projection refinement to enhance similarities between datasets for detection to perform alignment processing more efficiently ([0075-0077], Lukac).
Regarding claim 21, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the first electronic device of claim 17, wherein the one or more processors are further configured to execute the instructions to cause the first electronic device to identify the feature information by (as discussed above in claim 17): obtain the identification information according to an image processing task (page 1130, 2nd column, as discussed above of the obtaining of the metafeatures and the metric information, which is according to the classification task to optimize relevant machine learning framework [page 1131, 1st col, “ML Algorithm and Hyperparameters” section, 1st par.], by BRI, is analogous to the claimed limitation).
Regarding claim 23, Fuerer discloses a computer program product comprising instructions that are stored on a computer-readable medium and that, when executed by one or more processors, cause a first electronic device to (abstract discloses the processing is for machine learning which can be understood to include the use of a computer which includes a memory storing instructions to be executed by a processor): processing, by the one or more processors of the first electronic device, using at least one of the plurality of feature extraction network models, the to-be-processed image to extract feature information of a to-be-processed image (page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model] sine the claim simply recites “feature extraction network model” which doesn’t explicitly indicates a machine learning or a network of a neural network or not, therefore, by BRI, the examiner reads it to fall within the scope of Fuerer’s algorithm 2 wherein the metafeatures are calculated from the datasets [feature extraction] based on an algorithm with specific steps [network model]; the processor performing the function of algorithm 2 can be understood to be the first device as claimed; furthermore, page 1131 and FIG. 1 shows that the dataset includes dataset from OpenML and used for classification task such as liver disorders classification which can be understood to include dataset of images [any can be understood to the recited “to-be-processed image”]); identify the feature information to obtain identification information of the feature information (page 1130, 2nd col., 1st par., discloses a metric is calculated for the dataset based on the metafeatures, therefore, the measure information used to obtain the metric is analogous to the identification information as claimed, by BRI), wherein the identification information of the feature information comprises a first identifier of the at least one of the plurality of feature extraction network models (the processing of metafeature computation and algorithm 2 to be the pre-stored feature extraction network model; moreover, the metric’s measures obtained from the algorithm 2 is the identification information discussed above in claim 1 as disclosed in page 1130, 2nd col.; therefore, the metric itself can be understood as the identifier as claimed by BR); and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information (page 1129 discloses algorithm 1to select the model optimal for performing the function most suitable for the dataset based on the metric’s measure information obtained and the metafeature which is analogous to the claimed limitation, the processor performing algorithm 1 can be understood to be the second device as claimed, by BRI).
However, Fuerer does not explicitly disclose at least one pre-stored feature extraction network model of a neural network; a feature analysis network model of the neural network; wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image.
In the same field of Meta Learning and Algorithm (title, Vanschoren) Vanschoren discloses at least one pre-stored feature extraction network model of a neural network; a feature analysis network model of the neural network (algorithm 2, and the selected models as disclosed in section 3, 1st 4 paragraphs, are analogous to the algorithm 2 and the selected model of Fuerer, moreover, Vanschoren further teaches that the algorithm 2 and the processing of the selection is for neural network task such as disclosed in section 1, 1st par. hence, is analogous to “of a neural network” as claimed, as further explained in section 2 of page 25).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer to perform extracting, by a first device and using at least one pre-stored feature extraction network model, feature information of a to-be-processed image; identifying, by the first device, the feature information to obtain identification information of the feature information; and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information; wherein the at least one pre-stored feature extraction network model is of a neural network and the feature analysis network model if of the neural network as taught by Vanschoren to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to apply machine learning to novel dataset in an improve and efficient way (abstract, Vanschoren).
However, Fuerer in view of Vanschoren does not explicitly disclose wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image.
In the same field of dataset similarity measurement (abstract and [0039], Lukac) Lukac discloses wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model ([0076] discloses a refinement process to use one or more of feature extraction, matching, corner detection, template matching, semantic segmentation [semantic segmentation network], object detection [target detection network], image registration, trained classifier [image classification network] to adaptively maximize similarities in the overlapping region in images from two neighboring cameras [which is used to maximize similarities between two datasets which is analogous to Fuerer’s page 1130, 2nd column, discloses a metafeature computation from the input dataset, and further these metafeature of the dataset is calculated a matric that reflects how similar the datasets are [page 1130, 2nd column, 1st par.] using algorithm 2 [can be understood to be pre-stored feature extraction network model] sine the claim simply recites “feature extraction network model”], which Fuerer uses the algorithm to configure optimalization for similar dataset [page 1130, 1st par.]); receiving, from a sensor of the first device, a to-be-processed image ([0075] discloses the image are obtained from a camera of a device).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren to perform extracting, by a first device and using at least one pre-stored feature extraction network model, feature information of a to-be-processed image; identifying, by the first device, the feature information to obtain identification information of the feature information; and sending, by the first device and to a second device, the feature information and the identification information to indicate to the second device to select a feature analysis network model corresponding to the identification information to process the feature information; wherein the at least one pre-stored feature extraction network model is of a neural network and the feature analysis network model if of the neural network; wherein the plurality of feature extraction network models comprises an image classification network model, a target detection network model, and a semantic segmentation network model; receiving, from a sensor of the first device, a to-be-processed image as taught by Lukac to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to uses different models to perform projection refinement to enhance similarities between datasets for detection to perform alignment processing more efficiently ([0075-0077], Lukac).
Regarding claim 27, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the computer program product of claim 23, wherein the instructions, when executed by the one or more processors, further cause the first electronic device to (as discussed above in claim 23): obtaining the identification information according to an image processing task (page 1130, 2nd column, as discussed above of the obtaining of the metafeatures and the metric information, which is according to the classification task to optimize relevant machine learning framework [page 1131, 1st col, “ML Algorithm and Hyperparameters” section, 1st par.], by BRI, is analogous to the claimed limitation).
Regarding claim 28, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 1, wherein the identification information further comprises a second identifier of an output layer of the at least one of the plurality of feature extraction network models (page 1130, 2nd col., 1st par., discloses the metric calculated represent measure between the data, moreover, the measures are calculated for all data, such as shown in equation 5 therefore, any of which of these calculated measures can be understood to be the second identifier as claimed of the output of the calculation of algorithm 2 based on the metafeature [output layer of the feature information as claimed by BRI]), and wherein the output layer is a layer at which the feature information is output from the at least one of the plurality of feature extraction network models (as discussed previously, the measures are output of algorithm 2, in other words, is analogous to the claimed limitation wherein the algorithm 2 is analogous to the pre-stored feature extraction network model as claimed, as already discussed and explained in claim 1 above).
Regarding claim 30, Fuerer in view of Vanschoren and Lukac, wherein Lukac discloses the image processing method of claim 1, wherein the first device comprises a photographing apparatus, a camera, or a mobile terminal (“or” indicates a selection, the examiner selects “a camera” as disclosed in [0075] discloses the image are obtained from a camera of a device, of Lukac).
Regarding claim 31, Fuerer in view of Vanschoren and Lukac, wherein Lukac discloses the image processing method of claim 1, wherein sending the feature information and the identification information comprises wirelessly transmitting the feature information and the identification information from the first device to the second device (Lukac’s [0095] discloses wireless transmitting signals within communication channels to share data and coordinate movement within an autonomous driving environment, among other examples communications including sharing data of the cameras of these devices, which is analogous to transmitting feature information and identification information from one device to another).
Regarding claim 33, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 1, further comprising storing, in the memory of the first device, a plurality of feature analysis identifiers, wherein each feature analysis identifier corresponds to a different type of image processing task, and wherein the first identifier is selected based on a target image processing task to be performed on the to-be-processed image (page 1129, “Initializing SMBO With Configurations Suggested by Meta-Learning” section, discloses initializing with configurations by meta-learning, each configuration is initialized specifically for each meta learning performed by algorithm 2, therefore, different configurations are being output based on the different metalearning initializations, therefore, each initialization configuration can be understood to be an identifier for that specific suggestion by meta learning of algorithm 2).
Claims 4, 13, 16, 20, 22 and 26 are rejected under 35 U.S.C. 103 as being unpatentable over Matthias Feurer et. al. (“Initializing Bayesian Hyperparameter Optimization via Meta-Learning, 2015, Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 29, No. 1 [2015]: Twenty-Nonth AAAI Conference on Artificial Intelligence, AAAI Technical Track: Heuristic Search and Optimization” hereinafter as “Feurer”) in view of Joaquin Vanschoren et. al. (Meta-Learning and Algorithm Selection Workshop at ECAL, 2014, 21st European Conference on Artificial Intelligence” hereinafter as “Vanschoren”) further in view of Rastislav Lukac et. al. (“US 2019/0325580 A1” hereinafter as “Lukac”) and Yann N. Dauphin et. al. (“MetaInit: Initializing learning by learning to initialize, 2019, Advances in Neural Information Processing Systems 32, Curran Associates, Inc., pp. 12624-12636” hereinafter as “Dauphin”).
Regarding claim 4, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 1, wherein identifying the feature information comprises: obtaining the first identifier of the at least one of the plurality of feature extraction network models (as discussed above in claim 1, the metric obtained can be understood to be analogous to the first identifier as claimed, by BRI, which is the output of algorithm 2, in other words, is analogous to “first identifier of the at least one pre-stored feature extraction network model” as claimed); obtaining a second identifier of an output layer of the feature information (page 1130, 2nd col., 1st par., discloses the metric calculated represent measure between the data, moreover, the measures are calculated for all data, such as shown in equation 5 therefore, any of which of these calculated measures can be understood to be the second identifier as claimed of the output of the calculation of algorithm 2 based on the metafeature [output layer of the feature information as claimed by BRI]), wherein the output layer is a layer at which the feature information is output in the at least one of the plurality of feature extraction network models (as discussed previously, the measures are output of algorithm 2, in other words, is analogous to the claimed limitation wherein the algorithm 2 is analogous to the pre-stored feature extraction network model as claimed, as already discussed and explained in claim 1 above); and using the first identifier and the second identifier as the identification information of the feature information (the measures are used as metric is analogous to using the first identifier and the second identifier as the identification information of the metafeatures as claimed, by BRI, are analogous).
However, Fuerer in view of Vanschoren and Lukac does not explicitly disclose wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network.
In the same field of Meta Learning (title, Dauphin) Dauphin discloses wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network (“is one of…or…” indicates a selection, only one of these options is the instant scope of the claim, the examiner selects “a convolutional layer” which is taught in Dauphin’s page 7, “CIFAR” section which includes using convolutional layers to perform the metalearning initialization which is analogous to algorithm 2 of Fuerer, therefore, the it can be understood the output layer of algorithm 2 of Fuerer can be performed using a convolutional layer of Dauphin).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac to perform obtaining an identifier of an output layer of the feature information, wherein the output layer of the feature information is a layer at which the feature information is output in the at least one pre-stored feature extraction network model, and wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network as taught by Dauphin to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to perform meta learning initialization more effectively (abstract, Dauphin).
Regarding claim 13, Fuerer in view of Vanschoren and Lukac and Dauphin, wherein Fuerer discloses the image processing method of claim 4, wherein the identification information comprises a first field and a second field, and wherein the first field indicates the first identifier and the second field indicates the second identifier (page 1130, 2nd col., 1st par., discloses the measures are calculated in equation 5 for all of the data, therefore, any of which can be understood to be the first identifier as claimed and any of the others can be understood to be the second identifier as claimed, by BRI, moreover, in equation 5 the d(DN+1,Dj) can be understood to include the field from D-N+1 to Dj which can be understood to be the field as claimed, which as being respective to the first identifier and the second identifier’s metrics can be understood to be the first field and the second field, respectively, by BRI).
Regarding claim 16, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 14, wherein the identification information further comprises a second identifier of an output layer of the feature information, wherein the output layer is a layer at which the feature information is output in the at least one of the plurality of feature extraction network models (page 1130, 2nd col., 1st par., discloses the metric calculated represent measure between the data, moreover, the measures are calculated for all data, such as shown in equation 5 therefore, any of which of these calculated measures can be understood to be the second identifier as claimed of the output of the calculation of algorithm 2 based on the metafeature [output layer of the feature information as claimed by BRI of the feature network model]).
However, Fuerer in view of Vanschoren and Lukac does not explicitly disclose wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network.
In the same field of Meta Learning (title, Dauphin) Dauphin discloses wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network (“is one of…or…” indicates a selection, only one of these options is the instant scope of the claim, the examiner selects “a convolutional layer” which is taught in Dauphin’s page 7, “CIFAR” section which includes using convolutional layers to perform the metalearning initialization which is analogous to algorithm 2 of Fuerer, therefore, the it can be understood the output layer of algorithm 2 of Fuerer can be performed using a convolutional layer of Dauphin).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac to perform obtaining an identifier of an output layer of the feature information, wherein the output layer of the feature information is a layer at which the feature information is output in the at least one pre-stored feature extraction network model, and wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network as taught by Dauphin to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to perform meta learning initialization more effectively (abstract, Dauphin).
Regarding claim 20, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the first electronic device of claim 17, wherein the one or more processors are further configured to execute the instructions to cause the first electronic device to identify the feature information by: obtaining a first identifier of the at least one of the plurality of feature extraction network models (as discussed above in claim 17, the metric obtained can be understood to be analogous to the first identifier as claimed, by BRI, which is the output of algorithm 2, in other words, is analogous to “first identifier of the at least one pre-stored feature extraction network model” as claimed); obtaining a second identifier of an output layer (page 1130, 2nd col., 1st par., discloses the metric calculated represent measure between the data, moreover, the measures are calculated for all data, such as shown in equation 5 therefore, any of which of these calculated measures can be understood to be the second identifier as claimed of the output of the calculation of algorithm 2 based on the metafeature [output layer of the feature information as claimed by BRI]), wherein the output layer of the feature information is a layer at which the feature information is output in the at least one pre-stored feature extraction network model (as discussed previously, the measures are output of algorithm 2, in other words, is analogous to the claimed limitation wherein the algorithm 2 is analogous to the pre-stored feature extraction network model as claimed, as already discussed and explained in claim 17 above); and using the first identifier and the second identifier as the identification information of the feature information (the measures are used as metric is analogous to using the first identifier and the second identifier as the identification information of the metafeatures as claimed, by BRI, are analogous).
However, Fuerer in view of Vanschoren and Lukac does not explicitly disclose wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network.
In the same field of Meta Learning (title, Dauphin) Dauphin discloses wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network (“is one of…or…” indicates a selection, only one of these options is the instant scope of the claim, the examiner selects “a convolutional layer” which is taught in Dauphin’s page 7, “CIFAR” section which includes using convolutional layers to perform the metalearning initialization which is analogous to algorithm 2 of Fuerer, therefore, the it can be understood the output layer of algorithm 2 of Fuerer can be performed using a convolutional layer of Dauphin).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac to perform obtaining an identifier of an output layer of the feature information, wherein the output layer of the feature information is a layer at which the feature information is output in the at least one pre-stored feature extraction network model, and wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network as taught by Dauphin to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to perform meta learning initialization more effectively (abstract, Dauphin).
Regarding claim 22, Fuerer in view of Vanschoren and Lukac and Dauphin, wherein Fuerer discloses the first electronic device of claim 20, wherein the identification information comprises a first field and a second field, and wherein the first field indicates the first identifier and the second field indicates the second identifier (page 1130, 2nd col., 1st par., discloses the measures are calculated in equation 5 for all of the data, therefore, any of which can be understood to be the first identifier as claimed and any of the others can be understood to be the second identifier as claimed, by BRI, moreover, in equation 5 the d(DN+1,Dj) can be understood to include the field from D-N+1 to Dj which can be understood to be the field as claimed, which as being respective to the first identifier and the second identifier’s metrics can be understood to be the first field and the second field, respectively, by BRI).
Regarding claim 26, F Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the computer program product of claim 23, wherein the instructions, when executed by the one or more processors, further cause the first electronic device to: obtain the first identifier of the at least one of the plurality of feature extraction network models (as discussed above in claim 23, the metric obtained can be understood to be analogous to the first identifier as claimed, by BRI, which is the output of algorithm 2, in other words, is analogous to “first identifier of the at least one pre-stored feature extraction network model” as claimed); obtain a second identifier of an output layer of the feature information (page 1130, 2nd col., 1st par., discloses the metric calculated represent measure between the data, moreover, the measures are calculated for all data, such as shown in equation 5 therefore, any of which of these calculated measures can be understood to be the second identifier as claimed of the output of the calculation of algorithm 2 based on the metafeature [output layer of the feature information as claimed by BRI]), wherein the output layer is a layer at which the feature information is output in the at least one of the plurality of feature extraction network models (as discussed previously, the measures are output of algorithm 2, in other words, is analogous to the claimed limitation wherein the algorithm 2 is analogous to the pre-stored feature extraction network model as claimed, as already discussed and explained in claim 1 above); and use the first identifier and the second identifier as the identification information of the feature information (the measures are used as metric is analogous to using the first identifier and the second identifier as the identification information of the metafeatures as claimed, by BRI, are analogous).
However, Fuerer in view of Vanschoren and Lukac does not explicitly disclose wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network.
In the same field of Meta Learning (title, Dauphin) Dauphin discloses wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network (“is one of…or…” indicates a selection, only one of these options is the instant scope of the claim, the examiner selects “a convolutional layer” which is taught in Dauphin’s page 7, “CIFAR” section which includes using convolutional layers to perform the metalearning initialization which is analogous to algorithm 2 of Fuerer, therefore, the it can be understood the output layer of algorithm 2 of Fuerer can be performed using a convolutional layer of Dauphin).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac to perform obtaining an identifier of an output layer of the feature information, wherein the output layer of the feature information is a layer at which the feature information is output in the at least one pre-stored feature extraction network model, and wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network as taught by Dauphin to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to perform meta learning initialization more effectively (abstract, Dauphin).
Claim 29 is rejected under 35 U.S.C. 103 as being unpatentable over Matthias Feurer et. al. (“Initializing Bayesian Hyperparameter Optimization via Meta-Learning, 2015, Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 29, No. 1 [2015]: Twenty-Nonth AAAI Conference on Artificial Intelligence, AAAI Technical Track: Heuristic Search and Optimization” hereinafter as “Feurer”) in view of Joaquin Vanschoren et. al. (Meta-Learning and Algorithm Selection Workshop at ECAL, 2014, 21st European Conference on Artificial Intelligence” hereinafter as “Vanschoren”) further in view of Rastislav Lukac et. al. (“US 2019/0325580 A1” hereinafter as “Lukac”) and Yann N. Dauphin et. al. (“MetaInit: Initializing learning by learning to initialize, 2019, Advances in Neural Information Processing Systems 32, Curran Associates, Inc., pp. 12624-12636” hereinafter as “Dauphin”) and Lingqiao Liu et. al. (“Cross-Convolutional-Layer Pooling for Image Recognition, Nov. 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 39, Issue 11” hereinafter as “Liu”).
Regarding claim 29, Fuerer in view of Vanschoren and Lukac, wherein Fuerer discloses the image processing method of claim 28, wherein different output layers of the at least one of the plurality of feature extraction network models produce different feature information.
However, Fuerer in view of Vanschoren and Lukac does not explicitly disclose wherein each of the plurality of feature extraction network models comprises a plurality of layers including convolutional layers and pooling layers, wherein the output layer comprises an intermediate layer or a final layer of the at least one of the plurality of feature extraction network models (as discussed previously, the measures are output of algorithm 2, in other words, is analogous to the claimed limitation wherein the algorithm 2 is analogous to the pre-stored feature extraction network model as claimed, as already discussed and explained in claim 1 above).
In the same field of Meta Learning (title, Dauphin) Dauphin discloses wherein each of the plurality of feature extraction network models comprises a plurality of layers including convolutional layers, wherein the output layer comprises an intermediate layer or a final layer of the at least one of the plurality of feature extraction network models (“a convolutional layer” which is taught in Dauphin’s page 7, “CIFAR” section which includes using convolutional layers to perform the metalearning initialization which is analogous to algorithm 2 of Fuerer, therefore, the it can be understood the output layer of algorithm 2 of Fuerer can be performed using a convolutional layer of Dauphin; section 4.2, last par., discloses the each layer of the model, therefore, any last layer is analogous to the final layer since, “or” indicates a selection, and the examiner selects “final layer of the at least one of…”), and wherein different output layers of the at least one of the plurality of feature extraction network models produce different feature information (section 4.2, last par., discloses the different layers produce different feature information, including the output layers).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac to perform obtaining an identifier of an output layer of the feature information, wherein the output layer of the feature information is a layer at which the feature information is output in the at least one pre-stored feature extraction network model, and wherein the output layer is one of a convolutional layer, a pooling layer, or a fully connected layer of a neural network as taught by Dauphin to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to perform meta learning initialization more effectively (abstract, Dauphin).
However, Fuerer in view of Vanschoren and Lukac and Dauphin does not explicitly disclose feature extraction network models comprises a plurality of layers including convolutional layers and pooling layers.
In the same field of convolutional neural network processing (title and abstract, Liu) Liu discloses feature extraction network models comprises a plurality of layers including convolutional layers and pooling layers (section 3.1 discloses adding pooling for convolutional network to improve the functions of the convolutional neural network, to include pooling layers).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac and Dauphin to have a processing including a plurality of feature extraction wherein each of the plurality of feature extraction network models comprises a plurality of layers including convolutional layers and pooling layers as taught by Liu to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to perform convolutional neural network processing more efficiently (abstract, Liu).
Claim 32 is rejected under 35 U.S.C. 103 as being unpatentable over Matthias Feurer et. al. (“Initializing Bayesian Hyperparameter Optimization via Meta-Learning, 2015, Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 29, No. 1 [2015]: Twenty-Nonth AAAI Conference on Artificial Intelligence, AAAI Technical Track: Heuristic Search and Optimization” hereinafter as “Feurer”) in view of Joaquin Vanschoren et. al. (“Meta-Learning and Algorithm Selection Workshop at ECAL, 2014, 21st European Conference on Artificial Intelligence” hereinafter as “Vanschoren”) and Rastislav Lukac et. al. (“US 2019/0325580 A1” hereinafter as “Lukac”) and Montserrat Corbalan-Fuertes et. al. (“Color Image Acquisition by charge-coupled device cameras in polychromatic pattern recognition, March 1996, Optical Engineering, Vol. 35, Issue 3” hereinafter as “Corbalan-Fuertes”).
Regarding claim 32, Fuerer in view of Vanschoren and Lukac discloses the image processing method of claim 1 (as discussed above in claim 1).
However, Fuerer in view of Vanschoren and Lukac does not explicitly disclose wherein the sensor comprises a charge coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) photosensitive element.
In the same field of image recognition (title and abstract, Corbalan-Fuertes) Corbalan-Fuertes discloses wherein the sensor comprises a charge coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) photosensitive element (“or” indicates a selection, the examiner selects “charge coupled device” which is disclosed in section 3 wherein the camera used is a charge coupled device for recognition).
Thus, it would have been obvious for a person of ordinary skill in the art before the effective filing date to modify Fuerer in view of Vanschoren and Lukac to have a sensor camera to perform object recognition wherein the camera being a charge coupled device as taught by Corbalan-Fuertes to arrive at the claimed invention discussed above. Such a modification is the result of combing prior art elements according to known methods to yield predictable results. The motivation for the proposed modification would have been to uses different models to perform projection refinement to object recognition effectively using charge coupled cameras for pattern recognition and data processing robustly (abstract, Corbalan-Fuertes).
Pertinent Prior Art(s)
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Xin Li et. al., “US 2022/0076039 A1” discloses video-based activity recognition, data processing (abstract) discloses neural network for processing input image [0055] with different layers, using a second machine learning model for feature processing [0128-0131] and a CNN to analyze and learns the image [0094].
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PHUONG HAU CAI whose telephone number is (571)272-9424. The examiner can normally be reached M-F 8:30 am - 5:00pm.
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, Chineyere Wills-Burns can be reached at (571) 272-9752. 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.
/PHUONG HAU CAI/Examiner, Art Unit 2673
/CHINEYERE WILLS-BURNS/Supervisory Patent Examiner, Art Unit 2673
Prosecution Insights