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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 12/18/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
The information disclosure statement (IDS) submitted on 07/13/2026 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
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.
Claims 1, 2, 8-10 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Okanohara et al (Okanohara hereinafter US 10831577 B2) in view of Ko et al (Ko hereinafter US 20230290484 A1)
As per claim 1
Okanohara teaches An inspection device comprising: an interface device that receives data of an inspection target (Figure 1, Paragraph (18) “acquisition means for acquiring input data of abnormality detection target”) a storage device that stores an encoder model and a decoder model each of which is a machine learning model (Figure 1, Figure 2 Paragraph (11) “an encoder for inferring a latent variable from the input data based on the latent variable model stored in the storage means; a decoder for generating restored data from the latent variable based on the joint probability model stored in the storage means;”) a processor that is coupled to the interface device and the storage device (Figure 2) wherein the processor is configured to perform learning processing including learning the encoder model and the decoder model (Figure 7 Paragraph (32) “At this time, when learning the latent variable model, that is, the conditional probability p(z|x) from observation variable x to latent variable z using VAE in step 1, the learner is called an encoder. On the other hand, when learning the conditional probability p(x|z) from the latent variable z to the observation variable x, the learner is called a decoder.”) inference processing including inspecting the inspection target by using the learned encoder model and decoder model (Figure 7 Paragraph (63) “procedure 3 (realized by the measurement unit 12) learned encoders and decoders are prepared, and probability p(x) by P is calculated with determination data which performs abnormality detection as an input. “) the encoder model is a model that receives data of the inspection target as an input (Paragraph (33) “The input of the encoder is sensor data, obtained by the testing device” Paragraph (84) “ The data of abnormality detection target is input into the encoder of the learned model, the expression of the data of abnormality detection target is inferred”) outputs an abnormality degree of the inspection target (Paragraph (63) “The probability p(x) is correlated with the abnormality degree of each dimension with respect to the output result of the learned encoder and decoder with respect to the determination data input…probability p(x) is converted into score S(x)”)
In regards to the decoder model is a model in which an OK/NG value that indicates whether the inspection target is normal or abnormal and that is a value determined based on the abnormality degree and a feature amount of data of the inspection target are input, and in a case where the input OK/NG value indicates an abnormality, restored data of the inspection target is output based on the input feature amount.”
Okanohara states that “The probability p(x) is correlated with the abnormality degree of each dimension with respect to the output result of the learned encoder and decoder with respect to the determination data input…probability p(x) is converted into score S(x)” In Paragraph (63), “when S(x) exceeds the threshold value, it is determined as abnormality.” Therefore S(x) in relation to the threshold, allows for the claimed OK/NG value indicating whether the inspection target is normal or abnormal based on its abnormality degree (Okanohara’s p(x) probability). Furthermore, Okanohara also teaches the “feature amount” represented by expression z (which is the latent variable. Okanohara states that the encoder “infers the expression z (latent variable z) of the input data, inputs the expression z of the input data to the decoder, and generates the restored data x˜” in Paragraph (82). This teaches the claimed abnormality degree based normal/abnormal designation and separately teaches inputting the claimed feature amount to the decoder.
Okanohara does not expressly teach inputting the normal/abnormal designation to the decoder as a control input that governs whether decoder restoration is performed.
Ko also teaches obtaining a feature value from the inspection data and determining whether that data is normal or abnormal. Ko states in Paragraph [0069] “when the second medical data is determined to be abnormal, the analysis apparatus 100 may detect an abnormal area from the second medical data (S225), and generate the third medical data which is an image produced by restoring the abnormal area” Ko also states in paragraph [0071] “when the second medical data is determined to be a normal image, the data may be excluded.” Therefore, Ko teaches using a normal/abnormal designation to control whether restoration processing is performed. Restoration is performed for data designated abnormal while data designated normal may be excluded from restoration process.
In a combined system the inspection target data would first be processed according to Okanohara, to acquire feature amount “z” and abnormality degree s(x). The abnormality degree would be compared to Okanohara’s threshold to designate the inspection label as normal or abnormal. This value against the threshold is the claimed “OK/NG value that indicates whether the inspection target is normal or abnormal and that is a value determined based on the abnormality degree”. “latent variable z “is the claimed “a feature amount of data of the inspection target”. Then consistent with Ko’s teachings, the normal/abnormal designation would be modified into a control input for Okanohara’s decoder restoration processing along the latent variable z as the feature data input. When the abnormal/normal control input S(x) designates the target as normal restoration would not be performed. When S(x) control input designates abnormality, the decoder would do the restoration processing function using latent variable z and output restored data based on input variable z. The modified decoder receives both the abnormality degree based normal/abnormal value and the feature amount z. The normal/abnormal designation controls whether restoration occurs and the feature amount latent variable z provides the information which the restored data is generated.
Accordingly, a person of ordinary skill in the art, at the time this invention was effectively filed would have found it obvious to modify Okanohara’s abnormality detection system with Ko’s concept of gating/selectively controlling the restoration process of the decoder based on abnormal/normal designation value. Okanohara already determines whether the inspection target is abnormal based on its abnormality degree and possesses the latent feature amount used by its decoder for restoration. Meanwhile Ko teaches that the restoration processing may be selectively activated for data determined to be abnormal and ignored when it’s normal. Applying Ko’s selective restoration concept to Okanohara's already known normal/abnormal designation as a control input to the decoder along with the latent feature amount so that the decoder restoration is performed not when abnormal designation is found would have been predictable. A person of ordinary skill in the art would appreciate and have recognized that this modification reduces processing that isn’t necessary by the decoder’s restoration process for normal targets, saving restoration for specific anomalous designations.
As per claim 2
Okanohara and Ko teach all claim limitations previously rejected in claim 1’s 103 rejection. See claim 1’s 103 rejection.
Okanohara teaches in the inference processing, the processor outputs a value indicating whether an abnormality degree output from the encoder model is equal to or larger than a threshold as a value indicating whether the inspection target is normal or abnormal. (Paragraph (63) “earned encoders and decoders are prepared, and probability p(x) by P is calculated with determination data which performs abnormality detection as an input. The probability p(x) is correlated with the abnormality degree of each dimension with respect to the output result of the learned encoder and decoder with respect to the determination data input.” Paragraph (64) “In procedure 4 (realized by measurement unit 12 or determination unit 13), probability p(x) is converted into score S(x) to be smoothed. For example, conversion processing includes logarithm and the like.” Paragraph (67) “n procedure 5 (realized by the determination unit 13), when S(x) exceeds the threshold value, it is determined as abnormality.”)
As per claim 8
Okanohara and Ko teach all claim limitations previously rejected in claim 1’s 103 rejection. See claim 1’s 103 rejection.
Okanohara teaches wherein the encoder model is a neural network including a plurality of sequential intermediate layers, (Paragraph (27) “The neural network is not limited in its composition. For example, it may include a total coupling layer, a nonlinear function (sigmoid, relu), a batch normalization layer, a dropout layer”) in the learning processing, the feature amount input to the decoder model is a feature amount output from a last layer of the plurality of intermediate layers. Paragraph (89) “In this manner, in each layer of the decoder, processing of the decoder is performed by using parameters in the corresponding layer of the encoder…In this manner, the restored data x˜ generated in the decoder is compared with the training data x to calculate the difference, and the parameters of each layer of the encoder and decoder are updated so as to eliminate the difference. After that, learning is repeated so that the training data x and restored data x˜ coincide with each other.” Paragraph (94) “the training data x which is normal data is input into the encoder, the expression z (latent variable z) of the input data is inferred, the expression z of the input data is input into the decoder,”)
As per claim 9
Okanohara and Ko teach all claim limitations previously rejected in claim 1’s 103 rejection. See claim 1’s 103 rejection.
Okanohara teaches “wherein the interface device is coupled to a control device that controls one or a plurality of control target devices (Paragraph (118) “ the control unit such as the CPU 100 reads the program for each function of the monitoring control system stored in the auxiliary storage device”) ,the control device executes a control program and an information program (Paragraph (118) “the control unit executes the program read to the RAM 101, and the respective functional blocks of the present embodiment can be operated by one or a plurality of computers.” ) includes a shared memory for the control program and the information program (Figure 2, Figure 5, Figure 6 ) the control program is a program that performs scan processing (Paragraph (108) “ it can be used for home security monitoring system. It can also be used for monitoring systems such as buildings, public facilities, horticulture and the like.” Paragraph (118) “program of each function of the monitoring control system of the present embodiment “) which is processing for controlling a control target device and has real-time properties (Paragraph (101) “ In addition, for the determined action (occurred event), the actual action result (including the case where the action itself is not performed) actually performed by the related system is accumulated as feedback information, and the detection precision of the event is improved to make it possible to provide additional information that is highly usable for the system linked with the supervisory control system.” Paragraph (102) or example, it is possible to use for a failure prediction system of multiple industrial machines and robots in factories, and it is possible to use for an abnormality prediction system of infrastructure such as electric power. It is possible to use for an abnormality prediction system of multiple parts such as aircraft and automobiles.”) within a control cycle for each control cycle defined for the control program, (Paragraphs 101-117 give various examples where the abnormality detection system is used in systems that require control cycles) the information program is a program that performs information processing defined for the information program (Figure 5, Figure 6,) the shared memory includes a control area that is an area in which data is written from the control program (Figure 1, Figure 2) an information area that is an area, in which data is written from the information program the information area is smaller than the control area, (Figure 1, Figure 2 Allocation of memory capacity between the respective areas would have been an obvious design choice based on the amount of data required to be stored by the control and information functions. A skilled artisan would have recognized that allocating a smaller memory area to the information function than to the control function where the information function requires less storage would be an obvious way to conserve memory while providing adequate storage for each function. ) the processor outputs the OK/NG value to the control device through the interface device (Figure 1, Figure 2 Paragraph (16) “As shown in FIG. 2, the abnormality detection system 1 shown in FIG. 1 is configured as a computer system that physically includes a CPU 100, a RAM 101 which is a main storage device, and a ROM 102, an input/output device 103 such as a display, a communication module 104, and an auxiliary storage device 105, and the like. “ Paragraph (118) “the control unit such as the CPU 100 reads the program for each function of the monitoring control system stored in the auxiliary storage device 105 to the RAM 101, the control unit executes the program read to the RAM 101, and the respective functional blocks of the present embodiment can be operated by one or a plurality of computers. That is, one or a plurality of computers in which program of each function of the monitoring control system of the present embodiment is installed can operate as a computer device (system) performing each function alone or in cooperation.”) in the control device, the information program writes the OK/NG value in the information area, and the control program reads the OK/NG value from the information area and controls at least one control target device based on the OK/NG value. (Figure 1, Figure 2, Paragraph (118) “the control unit such as the CPU 100 reads the program for each function of the monitoring control system stored in the auxiliary storage device 105 to the RAM 101, the control unit executes the program read to the RAM 101, and the respective functional blocks of the present embodiment can be operated by one or a plurality of computers. That is, one or a plurality of computers in which program of each function of the monitoring control system of the present embodiment is installed can operate as a computer device (system) performing each function alone or in cooperation.”)
As per claim 10
Okanohara and Ko teach all claim limitations previously rejected in claim 1’s 103 rejection. See claim 1’s 103 rejection.
Okanohara wherein a learning parameter of the encoder model is different from a learning parameter of the decoder model learning parameter is a set of a parameter item and a parameter value for each learning parameter, and a feature in which the learning parameters are different means that parameter items are different and/or that parameter items are same, but parameter values are different. (Paragraph (32) “At this time, when learning the latent variable model, that is, the conditional probability p(z|x) from observation variable x to latent variable z using VAE in step 1, the learner is called an encoder. On the other hand, when learning the conditional probability p(x|z) from the latent variable z to the observation variable x, the learner is called a decoder.)
As per claim 12
Claim 12 is the parallel method claim to claim 1’s device claim therefore will be rejected under the same premise.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Okanohara et al (Okanohara hereinafter US 10831577 B2) in view of Ko et al (Ko hereinafter US 20230290484 A1) in further view of Bergmann et al (Bergmann hereinafter “The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection”)
As per claim 3
Okanohara and Ko teach all claim limitations previously rejected in claim 2’s 103 rejection. See claim 2’s 103 rejection.
Okanohara teaches wherein in the learning processing, each of a plurality of pieces of normal data is input to the encoder model, each of the plurality of pieces of normal data is data of a normal target (Figure 7, Paragraph (19) “a training data acquisition step of acquiring at least one or more pieces of training data consisting of normal data; an inference step of inferring a latent variable from the training data based on the latent variable model;”)
Okanohara discloses using a threshold in learning and inference (Paragraphs 40-67) but does not disclose how this threshold is established. Therefore, Okanohara nor Ko teach the processor determines a threshold of an abnormality degree on a basis of an index based on a variation in the abnormality degree output from the encoder model for each of the plurality of pieces of normal data.
Bergmann teaches the processor determines a threshold of an abnormality degree on a basis of an index based on a variation in the abnormality degree output from the encoder model for each of the plurality of pieces of normal data (Bergmann teaches determining an anomaly threshold using anomaly scores acquired from anomaly free data. Bergmann in section 6.1 Training and Evaluation Setup: “For each dataset category, we randomly split 10% of the anomaly-free training images into a validation set. The same validation set was used for all evaluated methods”. In section 5 Threshold Selection Bergmann states that their threshold k-sigma is “compute the mean μ and standard deviation σ over all anomaly scores of the validation set and then define a threshold to be t = μ+kσ” Bergmann states this calculation “takes the spread of the distribution of anomaly scores into account.”. This shows that Bergmann’s anomaly scores parallel with the claimed abnormality degrees. The standard deviation is an index based on the variation of those abnormality degrees and the threshold t = μ+kσ is based on that variation index.) and the determined threshold is the threshold in the inference processing. (Bergmann states in section 5 Threshold Selection that “the thresholds are estimated solely on a set of anomaly-free validation images prior to testing” then used during the testing disclosed above)
In the combined Okanohara/Ko/Bergmann system, Okanohara’s plurality of normal training data would be processed according to Okanohara’s encoder-based abnormality model to get the corresponding abnormality degrees. Bergmann’s threshold determination technique would then be applied to those abnormality degrees. Bergmann says to compute the mean μ and standard deviation σ over all anomaly scores of the validation set and then define a threshold to be t = μ+kσ”. The processor would determine the mean and standard deviation of the abnormality degrees generated for Okanohara’s normal data. Bergmann designates sigma as accounting for the “spread of the distribution of anomaly scores”. Therefore, sigma would provide the claimed index based on variation within Okanohara’s abnormality degrees. The processor would simply determine Okanohara’s abnormality threshold using Bergmann’s t = μ+kσ formulation. The threshold would be used as Okanohara’s inference threshold such that the abnormality degree S(x) of subsequently received inspection target data is compared with the determined threshold to designate the inspection target as normal or abnormal. Examiner notes Bergmann is only being relied upon for the manner in which Okanohara’s already disclosed abnormality threshold is actually determined.
Accordingly, a person of ordinary skill in the art at the time this invention was effectively filed would have found it obvious to further modify the Okanohara/Ko system with Bergmann’s known threshold determination technique. Okanohara’s abnormality threshold would be determined from the variation in abnormality degrees obtained from the normal data. Okanohara already processes normal data using their abnormality model and uses a threshold to determine whether processed data is abnormal while Bergmann teaches exactly how that threshold is established from anomaly free data. Bergmann expressly teaches that this technique accounts for the “spread of the distribution of anomaly scores” i.e. the variation. Applying Bergmann’s known statistical threshold technique to the previously modified Okanohara/Ko system would have been a predictable way of determining Okanohara’s already required inference threshold which is based on the variation of normal data. A person of ordinary skill in the art would recognize that this increases threshold reliability by accounting for variation in normal data. This reduces the plausibility of ordinary variations among normal inspections classified as abnormalities incorrectly.
Allowable Subject Matter
Claim 4-7 and 11 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
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHANE WRENSFORD CODRINGTON whose telephone number is (571)272-8130. The examiner can normally be reached 8:00am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Matthew Bella can be reached at (571) 272-7778. 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.
/SHANE WRENSFORD CODRINGTON/ Examiner, Art Unit 2667
/MATTHEW C BELLA/ Supervisory Patent Examiner, Art Unit 2667
Prosecution Insights