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 .
Response to Amendment
Applicant’s response received 12/31/2025 to the Non-Final Rejection mailed 10/9/2025 has been entered and made of record. Claims 1, 3-9, 11-15, and 17-20 remain pending.
Response to Arguments
Applicant's arguments filed 12/31/2025 have been fully considered but they are not persuasive. Applicant argues that the prior art does not disclose a preset number of normal images, which are representative of qualified products, are collected. Examiner respectfully disagrees, as the applicant is misconstruing the claim for what the specification actually says and supports. According to Paragraph [0015] of the Applicant’s originally filed specification, “The preset number of normal images represent images of a flawless product….The normal images are captured by using an industrial camera. Since images are generated by shooting products without defects under normal operation are normally flawless samples…” What this means to a person of ordinary skill in the art is that the images collected are by the flawless product, ie industrial camera, since it is shot without defects. This is the complete opposite of the amended portion being argued, which is that the present number of normal images representing images of flawless products, and is not the same as said in the specification disclosed directly above. As such the applicant’s arguments are not persuasive, and further, a 112(b) rejection is being added. Additionally, the arguments are directed solely to the amended portion that raises the 112(b) issue, and will not be considered, and the prior art rejection is maintained.
Based on these facts, the action is made FINAL.
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 1, 3-9, 11-15, and 17-20 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.
Regarding independent claims 1, 9, and 15, the claims disclose “the preset number of normal images representing images of flawless products…” This is unclear as Paragraph [0015] of the Applicant’s originally filed specification discloses, “The preset number of normal images represent images of a flawless product….The normal images are captured by using an industrial camera. Since images are generated by shooting products without defects under normal operation are normally flawless samples…” Thus, the amendment is unclear since it would appear to a person of ordinary skill in the art that the images collected are by the flawless product, ie industrial camera, as it is shot without defects. This is not the same as the present number of normal images representing images of flawless products. As such, the amendment is unclear and appropriate correction is required.
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.
Claims 1, 3-9, 11-15, and 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Jinwon An’s publication “Variational autoencoder based anomaly detection using reconstruction probability” (2015), hereinafter referred to as An, in view of Akilesh (WO 2019018693 A2), Hiroshi Takahashi’s publication “Student-t Variational Autoencoder for Robust Density Estimation” (2018), hereinafter referred to as Takahashi, Shi (CN 112149757 A, the attached English language translation is used hereinafter as the Official English language translation of this CN document, as applied in previous Office Action), and Guo, Y., Liao, W., Wang, Q., Yu, L., Ji, T. & Li, P. (2018). Multidimensional Time Series Anomaly Detection: A GRU-based Gaussian Mixture Variational Autoencoder Approach. Proceedings of The 10th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research. 95:97-112., hereinafter referred to as Guo.
Regarding claim 1, An teaches an abnormality detection method, the method comprising: training an autoencoder model using normal images (An, “only data with normal instances are used to train the autoencoder,” pg. 4); inputting an image to be detected into the autoencoder model, obtaining a reconstruction probability p of the image to be detected based on a multi-layer neural network of the autoencoder model (An, Algorithm 4, Fig. 3, reconstruction probability p of the image [input MNIST dataset consists of images, pg. 12] is obtained based on multi-layer neural network of the autoencoder model, pg. 9), and determining whether a reconstructed image can be generated based on the image to be detected using the autoencoder model according to the reconstruction probability and a reconstruction threshold (An, Algorithm 4, determine whether a reconstructed image is an anomaly [if it can be generated, pg. 2] according to the reconstruction probability, which is based on the image, and a reconstruction threshold, pg. 9), in response that no reconstructed image is generated based on the image to be detected, determining the image to be detected is an abnormal image (An, Algorithm 4, if reconstruction probability, based on the image to be detected and indicating whether the reconstruction image is generated, is below threshold, then it is an anomaly, pg. 9); or in response that the reconstructed image is generated based on the image to be detected, obtaining the reconstructed image to be detected (An, Algorithm 4, Fig. 4, if it is determined to not be an anomaly [can be/is generated, with low reconstruction probability] based on the image to be detected, then the reconstructed image is obtained, pgs. 9, 12, 14); and determining whether the reconstructed image is abnormal according to a defect judgment criterion (An, Algorithm 2, reconstructed image is determined to be abnormal/an anomaly according to reconstruction error [which is a defect judgement criterion], pg. 4).
However, An fails to teach where Akilesh teaches wherein training the autoencoder model using normal images comprises: collecting a preset number of normal images (Akilesh, “input image data,” [0019]); obtaining an implicit low dimensional vector h of each of the preset number of normal images in the preset number of normal images (Akilesh, obtain “vector T having reduced dimensionality,” for the data acquired from the input image data, [00139]).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified An using the teachings of Akilesh to include Akilesh’s compression of the image data into low dimensional vectors to An’s image data processing. Doing so would improve the image data processing by providing low-dimensional vectorization of the data, which would be used to retain the information while maximizing processing efficiency.
However, the combination of An and Akilesh fails to teach where Takahashi teaches training and learning a T distribution by using the implicit lowdimensional vector h of each of the preset number of normal images (Takahashi, train and learn “student t-distribution,” using “latent [low-dimensional] variable vectors,” pgs. 1, 4).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified An, as modified by Akilesh, using the teachings of Takahashi to include Takahashi’s training and learning a T distribution using low dimensional vectors of processed image data to An’s, as modified by Akilesh, distribution of low dimensional vectors of processed image data. Doing so would improve distribution of low dimensional vectors by providing T distribution, which would be used to make the distribution more robust to error (Takahashi, “Student-t distribution is…robust to the error, which makes the training stable,” pg. 6).
However, the combination of An, Akilesh, and Takahashi fails to teach where Shi teaches the reconstruction threshold being determined based on an expectation of a defect detection capability of the autoencoder model and reconstruction probabilities of the normal images (Shi, determine abnormal reconstruction error threshold [corresponding to reconstruction threshold] based on reconstruction error [corresponding to defect detection capability] and reconstruction error probability density distribution [corresponding to reconstruction probability] of the image, [0031-0035]).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified An, as modified by Akilesh and Takahashi, using the teachings of Shi to include Shi’s reconstruction threshold being determined based on an expectation of a defect detection capability of the autoencoder model and reconstruction probabilities of the normal images to An’s, as modified by Akilesh and Takahashi, reconstruction threshold. Doing so would improve the reconstruction threshold by providing consideration of the defect detection capability and reconstruction probabilities, which would be used to make the threshold more sensitive to the autoencoder’s capacity.
However, the combination of An, Akilesh, Takahashi, and Shi fails to teach where Guo teaches the reconstruction threshold being one value selected from a maximum value of reconstruction probabilities of the normal images and a preset percent quantile value of the reconstruction probabilities of the normal images (Guo, reconstruction threshold is selected and preset to be the “(100*r)-th” percent quantile value of the reconstruction probabilities of the dataset of normal images, pgs. 6, 9-10).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified An, as modified by Akilesh, Takahashi, and Shi, using the teachings of Guo to include Guo’s reconstruction threshold being a preset percent quantile value of the reconstruction probabilities to An’s, as modified by Akilesh, Takahashi, and Shi, reconstruction threshold. Doing so would improve the reconstruction threshold by providing a preset percent quantile value as the reconstruction threshold, which would be used to assess for anomalies more efficiently using a threshold dependent on the dataset.
Regarding claim 3, the combination of An, Shi, Guo, Akilesh, and Takahashi teach the abnormality detection method of claim 2, wherein obtaining the implicit low-dimensional vector h of each of the preset number of normal images comprises: performing an image processing on each of the preset number of normal images (Akilesh, takes the image input and processes the data, [00139]), and obtaining an image vector X corresponding to each of the preset number of normal images (Akilesh, obtain “vector S” for the data from the images [00139]); compressing the vector X of each of the preset number of normal images (Akilesh, “reduce the dimensionality…by applying a transformation to the vector S,” [00139]), and obtaining the implicit lowdimensional vector h corresponding to each of the preset number of normal images (Akilesh, “transformation may produce an output vector T having reduced dimensionality relative to the vector S,” [00139]).
Regarding claim 4, the combination of An, Shi, Guo, Akilesh, and Takahashi teach the abnormality detection method of claim 2, wherein training and learning the T distribution by using the implicit lowdimensional vector h of each of the preset number of normal images comprises: acquiring the T distribution according to machine learning process by training a variational autoencoder model using a distribution similarity of the implicit low-dimensional vector h (Takahashi, train the “Student-t Variational Autoencoder” using the distribution of “latent [low-dimensional] variable vectors,” pgs. 4-5).
Regarding claim 5, the combination of An, Shi, Guo, Akilesh, and Takahashi teach the abnormality detection method of claim 2, wherein a density function expression of the T distribution is:
PNG
media_image1.png
130
498
media_image1.png
Greyscale
PNG
media_image2.png
216
248
media_image2.png
Greyscale
(Takahashi, Formula 11, pg.4)
Regarding claim 6, the combination of An, Shi, and Guo teaches the abnormality detection method claim 1, wherein determining whether the reconstructed image can be generated based on the image to be detected using the autoencoder model comprises: obtaining the reconstruction probability p of the image to be detected based on the multi-layer neural network of the autoencoder model (An, Algorithm 4, Fig. 3, reconstruction probability p of the image [input MNIST dataset consists of images, pg. 12] is obtained based on multi-layer neural network of the autoencoder model, pg. 9); in response that the reconstruction probability p of the image to be detected is less than or equal to the reconstruction threshold, determining that the reconstructed image cannot be generated based on the image to be detected (An, Algorithm 4, if reconstruction probability is less than the reconstruction threshold, then it is an anomaly [cannot be generated], based on information from the image to be detected, pg. 9); or in response that the reconstruction probability of the image to be detected is greater than the reconstruction threshold, determining that the reconstructed image can be generated based on the image to be detected. (An, Algorithm 4, if reconstruction probability is greater than the reconstruction threshold, then it is not an anomaly [can be generated], based on information from the image to be detected, pg. 9).
Regarding claims 9-20, the rationale provided in the rejection of claims 1 and 3-6 are incorporated herein. In addition, the method of claims 1 and 3-6 corresponds to the electronic device of claims 9 and 11-14 (Shi, electronic device, processor, and storage device, [0048]) and the non-transitory storage medium of claims 15 and 17-20 (Shi, computer-readable storage medium, [0051], [0200]), and performs the steps disclosed herein.
Claims 7-8 is/are rejected under 35 U.S.C. 103 as being unpatentable over An in view of Akilesh, Takahashi, Shi, and Guo as applied to claim 1 above, further in view of Beggel (EP 3477553 A1).
Regarding claim 7, the combination of An, Akilesh, Takahashi, Shi, and Guo teaches the abnormality detection method of claim 1, wherein determining that the image to be detected is a normal or an abnormal image is based on reconstruction error (An, Algorithm 2, image is determined to be abnormal/an anomaly according to reconstruction, pg. 4).
However, the combination of An, Akilesh, Takahashi, Shi, and Guo fails to teach where Beggel teaches that the defect judgment criterion comprises: obtaining an image mean square error (MSE) between the image to be detected and the reconstructed image (Beggel, reconstruction is “mean squared error” over all pixels between the image to be detected and the reconstructed image, [0019-0020]); in response that the MSE is less than or equal to an error threshold r, determining that the image to be detected is a normal image; or in response that the MSE is greater than the error threshold r, determining that the image to be detected is an abnormal image (Beggel, obtain “threshold on reconstruction error [MSE] and consider all images that exceed the threshold as an anomaly,” [0020]).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified An, as modified by Akilesh, Takahashi, Shi, and Guo, using the teachings of Beggel to include Beggel’s use of MSE as the reconstruction error to An’s, as modified by Akilesh, Takahashi, Shi, and Guo, reconstruction error. Doing so would improve the reconstruction error by providing the MSE calculation, which would be used to place higher weight on larger errors (an effect of using MSE), thereby increasing the autoencoder’s sensitivity.
Regarding claim 8, the combination of An, Akilesh, Takahashi, Shi, Guo, Akilesh, and Beggel teaches the abnormality detection method of claim 7, wherein a calculation formula of the MSE is: MSE = (y – y’)^2, y is a pixel of the image to be detected, y’ is a pixel of the reconstructed image of the image to be detected and corresponds to the pixel y (Beggel, “mean squared error over all pixels” between the image to be detected and the reconstructed image, [0020]; additionally, the MSE equation and the reconstruction error being between the image to be detected and the reconstructed image are both well-known to one of ordinary skill in the art).
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 Vincent Rudolph whose telephone number is (571)272-8243. The examiner can normally be reached M-F 7:30 AM - 3:30 PM.
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.
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.
/VINCENT RUDOLPH/Supervisory Patent Examiner, Art Unit 2671
Prosecution Insights