DETAILED ACTION
Election/Restrictions
Applicant’s election without traverse of Species I, Claims 1-3, 10-18, and 19-20 in the reply filed on 12/19/2025 is acknowledged.
Claims 4-9 are withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected Species, there being no allowable generic or linking claim. Election was made without traverse in the reply filed on 12/19/2025.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 5/25/2022 and 9/13/2023 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner. However, it is noted that All Non-Patent Literature (NPL) citations need at least a month and year of publication: MPEP 609.04(a): The date of publication supplied must include at least the month and year of publication, except that the year of publication (without the month) will be accepted if the applicant points out in the information disclosure statement that the year of publication is sufficiently earlier than the effective U.S. filing date and any foreign priority date so that the particular month of publication is not in issue. NPL cited without at least the month and year of publication has been labeled with “no date available”.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. The limitation(s) is/are “a system comprising: means for processing…” in claim 20. Referring to the specifications as filed, the corresponding structure disclosed that performs this function includes a processing circuit (¶50) executing the training procedures described in ¶38-48.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 10-14 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Specifically, claim 10 recites “wherein a first image of the first subset comprises a first portion, processed by a reconstruction neural network”, claim 11 recites “feeding the masked image to the reconstruction neural network, to form the first portion as an output of the reconstruction neural network”, claim 12 recites “wherein the first image further comprises a second portion, processed by a super-resolution neural network”, claim 13 recites “feeding the noisy image to the super-resolution neural network, to form the second portion as an output of the super-resolution neural network”, and claim 14 recites that “a third portion, processed by a super-resolution neural network, and a fourth portion, processed by a reconstruction neural network; the third portion is diagonally opposed to the second portion; and the fourth portion is diagonally opposed to the first portion.”. However, the scope of these limitations is unclear.
The specifications at ¶46 describes a different and inconsistent configuration. As set forth in ¶46, the specification teaches “In each of the first portion 410 and the fourth portion 425, defects may be introduced by adding noise and processing the resulting noisy image with a super-resolution (“S.R.”) neural network” and “In each of the second portion 415 and the third portion 420, defects may be introduced by masking out portions of the image and processing the resulting image using a reconstruction (“Recon.”) neural network (the defects being the result of the reconstruction neural network's imperfect ability to reconstruct the lost portions of the image). Thus, the specifications consistently describes the first and four portions being processed by a super resolution neural network and the second and third portions being processed by a reconstruction neural network. In contrast, claims 10-14 requires the reverse assignment of processing types to the recited portions. The specifications does not describe this reverse configuration, and does not disclose applying super resolution processing to the second and third portions while applying reconstruction processing to the first and fourth portions as claimed. ¶46 indicates that, in some embodiments, defects may be introduced into only some portions or using one of the disclose methods, but this general statement does not provide support for the specific arrangement and diagonal relationships recited in claims 10-14.
Because the specifications provides the only description of how the recited “first”, “second”, “third”, and “fourth” portions are defined and used, and that description is inconsistent with the claimed assignments, it is unclear how these positional terms are to be interpreted. It is therefore unclear which portions of the image correspond to which processing operations, and the metes and bounds of the claims cannot be determined with reasonable certainty. Therefore, claims 10-14 are indefinite.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-3 and 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Bergmann et al (NPL: Uninformed Students: Student–Teacher Anomaly Detection with Discriminative Latent Embeddings) in view of Li et al (NPL: CutPaste: Self-Supervised Learning for Anomaly Detection and Localization).
Regarding claim 1, Bergmann discloses a method, comprising:
training a first neural network with a first set of images (pg. 1 abstract: We introduce a powerful student–teacher framework for the challenging problem of unsupervised anomaly detection. Student networks are trained to regress the output of a descriptive teacher network that was pretrained on a large dataset of patches from natural images),
wherein:
the first neural network comprises:
a first student neural network (Fig. 2 An ensemble of M student networks is trained to regress the output of the teacher on anomaly-free data), and
a first teacher neural network (Fig. 2 Input images are fed through a teacher network that densely extracts features for local image regions);
the training of the first neural network with the first set of images comprises:
training the first student neural network with the first set of images (pg. 4 3.2. Ensemble of Student Networks for Deep Anomaly Detection; Descriptors are extracted by applying T to each image in the dataset D. We then train an ensemble of M ≥ 1 randomly initialized student networks Si , i ∈ {1, . . . , M} that possess the identical network architecture as the teacher T. For an input image I, each student outputs its predictive distribution over the space of possible regression targets for each local image region p(r,c) centered at row r and column c);
and
the training of the first student neural network comprises training the first student neural network with a first cost function (Fig. 2 An ensemble of M student networks is trained to regress the output of the teacher on anomaly-free data. During inference, the students will yield increased regression errors e and predictive uncertainties v in pixels for which the receptive field covers anomalous regions), that:
for an image of the first set and not of the first subset, rewards similarity between a feature map of the first student neural network and a feature map of the first teacher neural network (Fig. 2 An ensemble of M student networks is trained to regress the output of the teacher on anomaly-free data; regression is made by minimizing the difference (error) between the feature maps, which rewards similarity for normal anomaly free data), and
Bergmann fails to teach where Li teaches the training of the first neural network with the first set of images comprises: introducing defects into a first subset of the first set of images (abstract: We learn representations by classifying normal data from the CutPaste, a simple data augmentation strategy that cuts an image patch and pastes at a random location of a large image; pg. 3 we generate the pseudo anomalies (CutPaste) during the training), and for an image of the first subset, rewards dissimilarity between a feature map of the first student neural network and a feature map of the first teacher neural network (abstract: We first learn self-supervised deep representations and then build a generative one-class classifier on learned representations; pg. 3 we define the training objective of the proposed self-supervised representation learning as follows: LCP = Ex∈X {CE(g(x), 0) + CE(g(CP(x)), 1}. where X is the set of normal data, CP(·) is a CutPaste augmentation and g is a binary classifier parameterized by deep networks. CE(·, ·) refers to a cross-entropy loss; wherein cross-entropy loss penalizes wrong predictions (e.g. rewards dissimilarity)).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of the training of the first neural network with the first set of images comprises: introducing defects into a first subset of the first set of images and for an image of the first subset, rewards dissimilarity between a feature map of the first student neural network and a feature map of the first teacher neural network from Li into the method as disclosed by Bergmann. The motivation for doing this is to improve methods for learning for anomaly detection and localization.
Regarding claim 2, the combination of Bergmann and Li disclose the method of claim 1, further comprising training the first teacher neural network with a second set of images and a second cost function (Bergmann pg. 3 3. Student–Teacher Anomaly Detection: a descriptive teacher network T pretrained on a large dataset of natural images; pg. 4 3.1. Learning Local Patch Descriptors: The final training loss for Tˆ is then given as L(Tˆ) = λkLk(Tˆ) + λmLm(Tˆ) + λcLc(Tˆ); pg. 6 4. Experiments: For the pretraining of the teacher networks Tˆ, triplets augmented from the ImageNet dataset are use), wherein: the second set of images comprises images each labeled with a classification label (Bergmann pg. 3 3. Student–Teacher Anomaly Detection: The teacher T has the same network architecture as the student networks. However, it remains constant and extracts descriptive embedding vectors for each pixel of the input image I that serve as deterministic regression targets during student training); and the second cost function rewards, for each image, similarity between a classification generated by the first teacher neural network and the classification label of the image (Bergmann pg. 4 3.1. Learning Local Patch Descriptors: The final training loss for Tˆ is then given as L(Tˆ) = λkLk(Tˆ) + λmLm(Tˆ) + λcLc(Tˆ) where λk, λm, λc ≥ 0 are weighting factors for the individual loss terms; because all components losses are minimization objectives, this teaches “rewards similarity”).
Regarding claim 3, the combination of Bergmann and Li disclose the method of claim 1, wherein the first neural network further comprises: a second student neural network (Bergmann Fig. 2 An ensemble of M student networks is trained to regress the output of the teacher on anomaly-free data), and a second teacher neural network (Bergmann Fig. 2 showing multiple teacher networks; pg. 6 4. Experiments: e.g. pretraining of the teacher networks T).
Regarding claim(s) 19 (drawn to a system):
The rejection/proposed combination of Bergmann and Li, explained in the rejection of method claim(s) 1, anticipates/renders obvious the steps of the system of claim(s) 19 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1 is/are equally applicable to claim(s) 19. See Bergmann pg. 1 Introduction: e.g. supervised computer vision algorithms, wherein a computer would have a CPU/Processor to run the student-teacher networks.
Regarding claim(s) 20 (drawn to a system):
The rejection/proposed combination of Bergmann and Li, explained in the rejection of method claim(s) 1, anticipates/renders obvious the steps of the system of claim(s) 20 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1 is/are equally applicable to claim(s) 20. See Bergmann pg. 1 Introduction: e.g. supervised computer vision algorithms, wherein a computer would have a CPU/Processor to run the student-teacher networks.
Claim(s) 10-11 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Bergmann and Li as applied to claim 1 above, and further in view of Mann et al (US Patent 11398255 B1).
Regarding claim 10, the combination of Bergmann and Li disclose the method of claim 1, but fail to teach where Mann teaches wherein a first image of the first subset comprises a first portion, processed by a reconstruction neural network (col 3 lines 55-35 training the machine learning model may include generating, for each said image frame containing said instance of the object, a respective attention mask highlighting one or more features of said instance of the object, and training the deep neural network to process the respective attention masks alongside the generated composite images to reconstruct the at least one frame of the isolated instance of the object).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein a first image of the first subset comprises a first portion, processed by a reconstruction neural network from Mann into the method as disclosed by the combination of Bergmann and Li. The motivation for doing this is to modifying objects or portions of objects within an image.
Regarding claim 11, the combination of Bergmann, Li, and Mann disclose the method of claim 10, further comprising generating the first portion, the generating of the first portion comprising: masking out a portion of a normal image to form a masked image; and feeding the masked image to the reconstruction neural network, to form the first portion as an output of the reconstruction neural network (Mann col 3 lines 55-35 training the machine learning model may include generating, for each said image frame containing said instance of the object, a respective attention mask highlighting one or more features of said instance of the object, and training the deep neural network to process the respective attention masks alongside the generated composite images to reconstruct the at least one frame of the isolated instance of the object). The motivation to combine the references is discussed above in the rejection for claim 10.
Claim(s) 12-13 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Bergmann, Li, and Mann as applied to claim 10 above, and further in view of Pottorff et al (US Patent 20220130013 B1).
Regarding claim 12, the combination of Bergmann, Li, and Mann disclose the method of claim 10, but fail to teach where Pottorff teaches wherein the first image further comprises a second portion, processed by a super-resolution neural network (¶54 such a dataset can be used to train one or more neural networks for real time rendering super resolution. In at least one embodiment, a synthetic dataset can include image with intentional rendering artifacts added, as may relate to noise, specular aliasing, shader aliasing, geometry aliasing, hidden objects, ghosting, or dithering. In at least one embodiment, this data can then be used to train a super resolution model for image reconstruction).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the first image further comprises a second portion, processed by a super-resolution neural network from Pottorff into the method as disclosed by the combination of Bergmann, Li, and Mann. The motivation for doing this is to train neural networks for generating high quality image and video content.
Regarding claim 13, the combination of Bergmann, Li, Mann, and Pottorff disclose the method of claim 12, further comprising generating the second portion, the generating of the second portion comprising: adding noise to a portion of a normal image to form a noisy image; and feeding the noisy image to the super-resolution neural network, to form the second portion as an output of the super-resolution neural network (Pottorff ¶54 such a dataset can be used to train one or more neural networks for real time rendering super resolution. In at least one embodiment, a synthetic dataset can include image with intentional rendering artifacts added, as may relate to noise, specular aliasing, shader aliasing, geometry aliasing, hidden objects, ghosting, or dithering. In at least one embodiment, this data can then be used to train a super resolution model for image reconstruction). The motivation to combine the references is discussed above in the rejection for claim 12.
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Bergmann and Li as applied to claim 1 above, and further in view of Jang et al (US 20080204066).
Regarding claim 15, the combination of Bergmann and Li disclose the method of claim 1, further comprising classifying, by the first neural network, a product image of an article in a manufacturing process as including a defect (Bergmann pg. 1 1. Introduction: In automated industrial inspection scenarios, it is often desirable to train models solely on a single class of anomaly-free images to segment defective regions during inference; see Fig. 2 Input images are fed through a teacher network that densely extracts features for local image regions & Fig. 5 examples of Anomaly detection at multiple scales).
The combination of Bergmann and Li fail to teach where Jang teaches removing the article from the manufacturing process (¶8 an electrical test is performed to screen defects generated in wafer manufacturing or assembling processes and to identify and remove the defective products).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of removing the article from the manufacturing process from Jang into the method as disclosed by combination of Bergmann and Li. The motivation for doing this is to improve method cost and efficiency.
Claim(s) 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Bergmann, Li, and Jang as applied to claim 15 above, and further in view of Huang et al (US 20220198181).
Regarding claim 16, the combination of Bergmann, Li, and Jang disclose the method of claim 15, wherein the classifying of the product image as including a defect comprises: feeding the product image to the first student neural network and to the first teacher neural network (Bergmann Fig. 2 input images are fed through a teacher network that densely extracts features for local image regions. An ensemble of M student networks is trained to regress the output of the teacher on anomaly-free data. During inference, the students will yield increased regression errors e and predictive uncertainties v in pixels for which the receptive field covers anomalous regions).
The combination of Bergmann, Li, and Jang fails to teach where Huang teaches determining that a measure of the difference between a latent feature vector of the first student neural network and a corresponding latent feature vector of the first teacher neural network exceeds a threshold (¶34 in response to the vector difference between the student tensor of the adjusted student model STM and a teacher tensor of the teacher model TTM is lower than the learning threshold; therefore a determination is made if it exceeds or is lower than the threshold).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of determining that a measure of the difference between a latent feature vector of the first student neural network and a corresponding latent feature vector of the first teacher neural network exceeds a threshold from Huang into the method as disclosed by combination of Bergmann, Li, and Jang. The motivation for doing this is to improve methods for object identification.
Regarding claim 17, the combination of Bergmann, Li, Jang and Huang disclose the method of claim 16, wherein the measure of the difference is an L2 norm of the difference (Li pg. 12 A.1. Experiment with ResNet18: A single cycle of cosine learning rate decay schedule [33] and L2 weight regularization with a coefficient of 0:00003 are used). The motivation to combine the references is discussed above in the rejection for claim 1.
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Bergmann, Li, and Jang as applied to claim 15 above, and further in view of Lee et al (US 20200249651).
Regarding claim 18, the combination of Bergmann, Li, and Jang disclose the method of claim 15, but fail to teach where Lee teaches wherein the product image is an image of a display panel in a manufacturing flow (¶42 For example, the classifier 112 may use a deep neural network (DNN) for identifying patterns or features of a 2D image. In the present example, such patterns may be correlated to anomalies, faults, and defects that may occur in a manufacturing process of an electronic device (e.g., a glass panel of a display device)).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the product image is an image of a display panel in a manufacturing flow from Lee into the method as disclosed by combination of Bergmann, Li, and Jang. The motivation for doing this is to improve methods for detecting a fault.
Examiner’s Comments
The Office has established rejections under 35 U.S.C. 112(b) with regard to claims 10-14. The scope of claims 10-14 cannot be determined because of the identified issues presented above. The rejection to claims 10-14 under 35 U.S.C. 112(b) renders the applicant's claims as being incomprehensible as to preclude a reasonably detailed search of the prior art by the examiner. The examiner has attempted to identify all grounds for rejection under 35 U.S.C. 112(b). The examiner suggests that the applicant carefully review the claims in order to fix any and all issues that have and have not been highlighted by this office action.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN KY whose telephone number is (571)272-7648. The examiner can normally be reached Monday-Friday 9-5PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Vincent Rudolph can be reached at 571-272-8243. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KEVIN KY/Primary Examiner, Art Unit 2671
Prosecution Insights