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 .
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1, 4, 6-16, 20 and 21is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Darvish (WO 2021/232149 A1).
-Regarding claim 1, Darvish discloses a computer implemented method for defect detection in an imaging dataset of an object comprising integrated circuit patterns (A computer-implemented method, an inspection system and a non-transitory storage medium are provided, for automatically generating defect classification models, using machine learning, for the inspection of semiconductor and/or Printed Circuit Board (PCB) parts, abstract), the method comprising:- obtaining defect candidates in the imaging dataset (the inspection images are images of semiconductor or PCB parts, paragraph 68);- subsequently carrying out at least two stages (the subset of the inspection images used for training the multi-class classifier consists of a subset of the first subset, i.e. this second subset comprises the inspection images classified as 'defective' by the binary classifier, paragraph 93), each stage comprising the following steps: applying a stage specific defect detection method to the defect candidates (using a first subset of the inspection images, to determine whether the inspection images correspond to non defective or defective parts, paragraph 6); discarding defect-free defect candidates (the defect classification application 618 is used to classify or label the inspection images, with label information indicative of whether the part is defective or not, and if defective, with label information indicative of the defect type, paragraph 66); and- obtaining detected defects in the imaging dataset from the remaining defect candidates (The defect classification model is thus built from the trained binary and multi-class classifier." After both stages, the final set of detected defects (identified defect types) is obtained from the remaining candidates (those that passed both classifiers), paragraph 97).
-Regarding claim 4, Darvish further discloses for at least 50% of stages, the spatial extent of the defect candidates examined by the stage specific defect detection method is smaller than the spatial extent of defect candidates examined by the stage specific defect detection method in a preceding stage (the subset of the inspection images used for training the multi-class classifier consists of a subset of the first subset, i.e. this second subset comprises the inspection images classified as 'defective' by the binary classifier, paragraph 93; [Note: Darvish teaches that stage 2 (multi-class) operates on a strictly smaller set of candidates than stage 1 (binary), satisfying the "smaller spatial extent" limitation for ≥50% of stages]).
-Regarding claim 6, Darvish further discloses for at least 50% of stages, the computation time of the stage specific defect detection method on all defect candidates in that stage is lower than the computation time of the stage specific defect detection method on the defect-free defect candidates of the subsequent stage (most of the inspection images generated by optical inspection systems are labelled or associated with a non-defective or non-defective label (or class). A small portion of the inspection images, such typically between 0.01% to 10%, as an example only, correspond to defective parts, paragraph 67).
-Regarding claim 7, Darvish further discloses the false negative rate of the stage specific detection methods is below 5% (accuracy of the classifiers reach a given accuracy threshold (such as above 95%) in a given number of epochs, paragraph 97).
-Regarding claim 8, Darvish further discloses for at least 50% of stages, the precision of the stage specific defect detection method is higher than the precision of the stage specific defect detection method in a preceding stage (Different runs were performed on the same inspection images, with the training system always selecting the ResNet34 model and SGD optimizer for the binary classifier, and selecting deeper models such as ResNet152 and InceptionResNet combined to the SGD optimizer for the multi-class classifier, thereby confirming the accuracy of the system in selecting the best combinations with respect to classification accuracy for both the binary and the multi-class classifiers, paragraph 108).
-Regarding claim 9, Darvish further discloses at least two of the stage specific defect detection methods only differ in their control parameters (the first and second combinations of neural network architecture and optimizer differ for the binary and multi-class classifiers, paragraph 82).
-Regarding claim 10, Darvish further discloses less than 10% of the imaging dataset comprise a defect (A small portion of the inspection images, such typically between 0.01% to 10%, paragraph 67).
-Regarding claim 11, Darvish further discloses intermediate results obtained by a stage specific defect detection method are re-used by a stage specific defect detection method in a subsequent stage (the subset of the inspection images used for training the multi-class classifier consists of a subset of the first subset, i.e. this second subset comprises the inspection images classified as 'defective' by the binary classifier, paragraph 93).
-Regarding claim 12, Darvish further discloses one or more stage specific defect detection methods estimate properties of the defect candidates, and wherein the properties are used to discard defect-free defect candidates (using a first subset of the inspection images, to determine whether the inspection images correspond to non defective or defective parts, paragraph 6).
-Regarding claim 13, Darvish further discloses at least one stage specific defect detection method comprises a machine learning method, the machine learning method comprising a trained machine learning model .
-Regarding claim 14, Darvish further discloses at least 50% of stage specific defect detection methods comprise a machine learning method, and wherein, for these stages the complexity of the machine learning model of the machine learning method is higher than the complexity of the machine learning model of the machine learning method of a preceding stage (the system always selecting the ResNet34 model and SGD optimizer for the binary classifier, and selecting deeper models such as ResNet152 and InceptionResNet combined to the SGD optimizer for the multi-class classifier, paragraph 108).
-Regarding claim 15, Darvish further discloses for at least one stage, the stage specific defect detection method is selected from a set of stage specific defect detection methods (the proposed system and method can automatically select, from a list of existing machine learning (ML)-models, the most appropriate model to be used, and tuning of hyperparameters associated to said model can be realized using a simple grid search technique, paragraph 55).
-Regarding claim 16, Darvish further discloses for at least one stage, the stage specific defect detection method is automatically selected from the set of stage specific defect detection methods according to a predetermined criterion (automatically select, paragraph 55).
-Regarding claim 20, Darvish further discloses a computer-readable medium, on which a computer program executable by a computing device is stored, the computer program comprising code for executing a method of claim 1 (a non-transitory storage medium, paragraph 40).
-Regarding claim 21, Darvish further discloses a computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out a method of claim 1 (a non-transitory storage medium … having stored thereon computer-readable instructions for causing a processor, paragraph 40).
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.
Claim(s) 2, 3, 5 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Darvish (WO 2021/232149 A1) in view of Suni (Sunil, K. V., et al. “MVAi: A Framework and Three Stage Approach to Defect Detection.” 2021 IEEE Mysore Sub Section International Conference (MysuruCon). IEEE, 2021. (Year: 2021)).
-Regarding claim 2, Darvish is silent to teaching that at least three stages are carried out. However, the claimed limitation is well known in the art as evidenced by Suni.
In the same field of endeavor, Suni teaches at least three stages are carried out (see abstract).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Darvish with the teaching of Suni in order to improve classification features.
-Regarding claim 3, the combination further discloses three stages are carried out (Suni, abstract).
-Regarding claim 5, the combination further discloses the defect candidates comprise images in a first stage, bounding boxes within the images in a subsequent second stage and subsets of the bounding boxes in a subsequent third stage (Suni, The AE-SSIM model results for a different texture and object dataset are shown in Figure 4).
-Regarding claim 18, the combination further discloses one or more stage specific defect detection methods use a reference dataset to discard defect-free defect candidates (Suni, The testing set combined with the defect-free images in the test split was used to determine and report metrics, section 1).
Claim(s) 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Darvish (WO 2021/232149 A1) in view of Vered (U.S. PG-PUB NO. 2025/0045904).
-Regarding claim 17, Darvish is silent to teaching that for at least one stage, the stage specific defect detection method is selected from the set of stage specific defect detection methods using meta-data concerning the imaging dataset. However, the claimed limitation is well known in the art as evidenced by Vered.
In the same field of endeavor, Vered teaches for at least one stage, the stage specific defect detection method is selected from the set of stage specific defect detection methods using meta-data concerning the imaging dataset (metadata, paragraph 62).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Darvish with the teaching of Vered in order to provide useful information on adjustments that may be needed to the manufacturing tools, equipment, and/or compositions, depending on the type of errors identified, so as to promote higher yield.
Claim(s) 19 and 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Darvish (WO 2021/232149 A1) in view of Basistyy (U.S. PG-PUB NO. 2023/0012917).
-Regarding claim 19, Darvish further discloses the computer implemented method for defect detection comprises a machine learning model (neural network, paragraph 60; trained model, paragraph 78),
Darvish is silent to teaching that the stage specific defect detection method in at least 50% of stages comprises at least one intermediate layer followed by a classifier that classifies intermediate results as defect or non-defect in the machine learning model. However, the claimed limitation is well known in the art as evidenced by Basistyy.
In the same field of endeavor, Basistyy teaches the stage specific defect detection method in at least 50% of stages comprises at least one intermediate layer followed by a classifier that classifies intermediate results as defect or non-defect in the machine learning model (The decision layer may output a decision regarding the presence or absence of a defect inside the respective segment, paragraph 35).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Darvish with the teaching of Basistyy in order to provide detection precision is obvious to try with predictable results.
-Regarding claim 22, the combination further discloses a system for defect detection in an object comprising integrated circuit patterns (Darvish, PCB, paragraph 32), the system comprising: - an imaging device configured to provide an imaging dataset of the object comprising integrated circuit patterns (Basistyy, detector 112, paragraph 21); - one or more processing devices (Darvish, processor, paragraph 32); - one or more machine-readable hardware storage devices comprising instructions that are executable by one or more processing devices to perform operations comprising the method of claim 1 (Darvish, a non-transitory storage medium … having stored thereon computer-readable instructions for causing a processor, paragraph 40).
Response to Arguments
Applicant's arguments filed 8/11/2026 have been fully considered but they are not persuasive.
Regarding Claim Rejections - 35 USC § 102
Claim 1
Applicant argues that Darvish’s multi-class classifier does not discard defect-free defect candidates, because its inputs are images the binary classifier has already determined to be defective. Applicant further argues that Darvish lacks a second stage that again distinguishes defect-free from potentially defective candidates and discards the former.
The Examiner respectfully disagrees. Claim 1 requires that each stage comprise “applying a stage specific defect detection method to the defect candidates” and “discarding defect-free defect candidates.” The claim does not require that a later stage again distinguish defect-free candidates from potentially defective ones. In Darvish, the binary classifier is applied to the inspection images, and the images determined to be non-defective are discarded (paragraphs 6, 78). The candidates passed to the multi-class classifier are limited to the images classified as defective, so the defect-free candidates are considered as discarded from the set to which the second-stage method is applied (paragraphs 75, 93). The multi-class classifier is then applied to those remaining candidates, and the detected defects are obtained from them (paragraphs 7, 97). Darvish therefore discloses each stage as claimed. The rejection is maintained.
Claim 4
Applicant argues that Darvish’s subset of images has fewer images but that each image keeps the same spatial extent. Applicant contends that a reduction in the number of images is not a reduction in spatial extent, and points to the example at page 22, lines 11–16 of the specification.
The Examiner respectfully disagrees. Claim 4 recites “the spatial extent of the defect candidates examined,” referring to the candidates collectively rather than to each individual candidate. Under the broadest reasonable interpretation, this includes the overall region of the object covered by the candidates examined in a stage. Darvish’s second stage examines only the subset of images classified as defective (paragraph 93). That subset covers a smaller region of the inspected object than the full set examined in the first stage. The specification example cited by Applicant is not recited in the claim, and limitations from the specification are not read into the claims. The rejection is maintained.
Claim 6
Applicant argues that Darvish discloses that only a small portion of images correspond to defective parts (paragraph 67), but does not mention computation time as claimed.
The Examiner respectfully disagrees. Darvish need not use the words “computation time” where the arrangement it discloses necessarily results in the claimed relationship. Darvish discloses a ResNet34 architecture for the first-stage binary classifier and deeper models, such as ResNet152 and InceptionResNet, for the second-stage multi-class classifier (paragraphs 82, 108). Darvish also discloses that defect-free images make up the large majority of the candidates (paragraph 67). Applying the shallower first-stage method to all candidates therefore requires less computation than applying the deeper second-stage method to the defect-free candidates. The rejection is maintained.
Claim Rejections - 35 USC § 103
Claim 1
Applicant argues that Sunil does not disclose the feature of claim 1 that is missing from Darvish. As explained above, Darvish discloses every limitation of claim 1. Sunil is not relied upon for claim 1, but only for the additional limitations of the dependent claims. The argument is therefore not persuasive.
Claim 2
Applicant argues that neither Darvish nor Sunil discloses a third stage that includes discarding defect-free defect candidates.
The examiner respectfully disagrees. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). In this case, Darvish teaches stages that each apply a stage-specific detection method and discard defect-free candidates. Sunil teaches carrying out defect detection in three stages (Sunil, abstract). The combined teachings would have suggested carrying out at least three such stages. The rejection is maintained.
Claim 3
Applicant presents the same argument as for claim 2. For the same reasons given for claim 2, the combination of Darvish and Sunil teaches carrying out three stages as claimed. The rejection is maintained.
Claim 5
Applicant argues that Sunil Fig. 4 shows the AE-SSIM model results for a texture and object dataset, but does not show bounding boxes or subsets of bounding boxes.
The Examiner respectfully disagrees. Under the broadest reasonable interpretation, a “bounding box” is a region of an image that encloses a defect candidate, and a “subset” of a bounding box is a smaller region within it. Sunil Fig. 4 shows the model operating on full images and producing progressively localized defect regions within those images. The rejection is maintained.
Claim 17
Applicant argues that Vered states only that images can include associated metadata (paragraph 62), without describing how the metadata is used. Applicant further argues that Vered does not disclose selecting the stage-specific defect detection method using the metadata, and that Vered does not cure the deficiency of Darvish as to claim 1.
These arguments are not persuasive. As explained above, Darvish discloses every limitation of claim 1. The remaining arguments attack Vered individually, whereas the rejection is based on the combination. Darvish teaches selecting the defect classification model, which comprises the stage-specific methods, according to the part type of the images being inspected (paragraph 101). Vered teaches that inspection images carry associated metadata (paragraph 62). Together, these teachings suggest selecting the stage-specific method using metadata concerning the imaging dataset. The rejection is maintained.
Claims 19 and 22
Applicant argues only that Basistyy does not cure the deficiency of Darvish as to claim 1. As explained above, Darvish discloses every limitation of claim 1. The argument is not persuasive, and the rejection is maintained.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PING Y HSIEH whose telephone number is (571)270-3011. The examiner can normally be reached Monday-Friday, 9am-4pm.
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, Jennifer Mehmood can be reached at (571) 272-2976. 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.
/PING Y HSIEH/ Primary Examiner, Art Unit 2664
Prosecution Insights