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 § 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 and 5 is/are rejected under 35 U.S.C. 103 as being unpatentable over a non-patent literature by Wang, Xueqiu, Huanbing Gao, Zemeng Jia, and Zijian Li, entitled "BL-YOLOv8: An Improved Road Defect Detection Model Based on YOLOv8" and published in 10/10/2023 (hereinafter Wang) in view of CN patent application publication no. 116797873 to Wang et al. (hereinafter Wang2) and us patent application publication no. 2023/0017202 to Zhang.
For claim 1, Wang as applied teaches a method for detecting defects in a hydro turbine top cover based on an improved you-only-look-once version 8 (YOLOv8) model (see, e.g., abstract, which teaches a defect detection model based on YOLOv8), comprising:
S1, processing hydro turbine top cover defect images which are collected to obtain a hydro turbine top cover defect data set (see, e.g., sec. 4.2, which teaches preparing defect image dataset);
S2, dividing the hydro turbine top cover defect data set into a train set, a validation set and a test set based on a ratio of 6:2:2 (see, e.g., sec. 4.2, which teaches dividing the dataset into training, validation, and test sets);
S3, constructing a network model for detecting the defects in the hydro turbine top cover based on YOLOv8-convolutional block attention module (YOLOv8-CBAM) (see, e.g., sec. 3.3 and 4.4, which teach utilizing a CBAM module in YOLOv8);
S4, training the network model for detecting the defects in the hydro turbine top cover based on the YOLOv8-CBAM (see, e.g., sec. 4.1-4.2, which training YOLOv8), specifically comprising:
inputting the hydro turbine top cover defect data set in the step S1 into the network model for training (see, e.g., sec. 4.1-4.2, which training using the prepared dataset), wherein a total loss function of the network model comprises:
a binary cross-entropy (BCE) loss, a distance-weighted flight (DFL) loss, and a complete intersection over union (CIOU) loss (see, e.g., sec. 2, which teaches that YOLOv8 has BCE loss, DFL loss and CIOU loss), and a calculation formula of the total loss function is as follows:
LOSS = λ1LOSSBCE + λ2LOSSDFL + λ3LOSSCIOU where LOSSBCE represents a classification loss, LOSSDFL represents a localization loss, LOSSCIOU represents another localization loss (see, e.g., sec. 2, which teaches that BCE loss is classification loss and DFL and CIOU losses are regression loss), λ1 represents a weight of the BCE loss in a total loss for the classification loss, λ2 represents a weight of the DFL loss in the total loss for the localization loss, and λ3 represents a weight of the CIOU loss in the total loss for the another localization loss (see, e.g., sec. 2, which teaches that the loss calculation is a weighted combination of BCE, DFL and CIOU losses);
S5, generating the improved YOLOv8 model after the step S4 (see, e.g., sec. 4.5 and FIG. 9, which teach generating the enhanced YOLOv8 model); and
S6, detecting the defects in the hydro turbine top cover based on the improved YOLOv8 model (see, e.g., sec. 4.7-4.8, which teach detecting defects using the enhanced model).
While Wang as applied detecting defects on the road, it does not explicitly teach that the defects are in the hydro turbine top cover. Wang 2 in the analogous art teach detecting a fault associated with a top cover of a water turbine using a convolution neural network (see, e.g., lines 23-32 on page 2 of Wang2).
It would have been obvious to use Wang’s method to detect a water turbine fault as taught by Wang 2 because doing so would effectively improve the safety and reliability of the water turbine (see, e.g., lines 8-14 on page 2 of Wang2).
While Wang as applied does not explicitly teach, Zhang in the analogous art teaches dividing the input dataset into a train set, a validation set and a test set based on a ratio of 6:2:2 (see, e.g., par. 50 of Zhang).
It would have been obvious to divide the input dataset as taught by Zhang because doing so would suitably train the model (see par. 50 of Zhang).
For claim 5, Wang in view of Wang 2 and Zhang teaches that in the step S3, the network model for detecting the defects in the hydro turbine top cover based on the YOLOv8-CBAM comprises: a backbone network, a neck network and a head network (see, e.g., sec. 2 and 3 of Wang, which teach using a YOLOv8 having a backbone-neck-head network structure).
Claim(s) 2 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Wang2 and Zhang and further in view of us patent no. 11227187 to Weinberger.
For claim 2, while Wang in view of Wang2 and Zhang does not explicitly teach, Weinberger in the analogous art teaches that in the step S1, the processing hydro turbine top cover defect images comprises:
cropping the hydro turbine top cover defect images which are taken on-site to obtain cropped images (see, e.g., lines 47-67 in col. 8 and lines 1-10 in col. 9 of Weinberger, which teach synthetically altering, e.g., cropping, images for anomaly detection), and
expanding the cropped images by image enhancement techniques to obtain the hydro turbine top cover defect data set (see, e.g., lines 47-67 in col. 8 and lines 1-10 in col. 9 of Weinberger, which teach synthetically altering, e.g., zooming, images to enhance flaws or defects depicted therein).
It would have been obvious to enhance the defect images as taught by Weinberger because doing so would yield predictable results of generating more realistic images and more accurately trained model (see MPEP 2143(I)(D) and lines 47-67 in col. 8 and lines 1-10 in col. 9 of Weinberger).
Claim(s) 17 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Wang2, Weinberger and Zhang.
For claim 17, Wang as applied teaches a method for detecting defects in a hydro turbine top cover based on an improved YOLOv8 model (see, e.g., abstract, which teaches a defect detection model based on YOLOv8), comprising:
S1, collecting hydro turbine top cover defect images, and processing the hydro turbine top cover defect images to obtain a hydro turbine top cover defect data set (see, e.g., sec. 4.2, which teaches preparing defect image dataset);, specifically comprising:
cropping the hydro turbine top cover defect images to obtain cropped images, and
expanding the cropped images by image enhancement techniques to obtain the hydro turbine top cover defect data set;
S2, dividing the hydro turbine top cover defect data set into a train set, a validation set and a test set based on a ratio of 6:2:2 (see, e.g., sec. 4.2, which teaches dividing the dataset into training, validation, and test sets);
S3, constructing a network for detecting the defects in the hydro turbine top cover based on YOLOv8-CBAM (see, e.g., sec. 3.3 and 4.4, which teach utilizing a CBAM module in YOLOv8), wherein the network for detecting the defects in the hydro turbine top cover based on the YOLOv8-CBAM comprises: a backbone network, a neck network and a head network (see, e.g., sec. 2 and 3 of Wang, which teach using a YOLOv8 having a backbone-neck-head network structure);
S4, training the network for detecting the defects in the hydro turbine top cover based on the YOLOv8-CBAM to obtain a YOLOv8-CBAM model as the improved YOLOv8 model, specifically comprising:
inputting the hydro turbine top cover defect data set into the network for training (see, e.g., sec. 4.1-4.2, which teach training using the prepared dataset), wherein a total loss function of the YOLOv8-CBAM model comprises:
a BCE loss, a DFL loss, and a CIOU loss (see, e.g., sec. 2, which teaches that YOLOv8 has BCE loss, DFL loss and CIOU loss), and a calculation formula of the total loss function is as follows:
LOSS = λ1LOSSBCE + λ2LOSSDFL + λ3LOSSCIOU , where LOSSBCE represents a classification loss, LOSSDFL represents a localization loss, LOSSCIOU represents another localization loss (see, e.g., sec. 2, which teaches that BCE loss is classification loss and DFL and CIOU losses are regression loss), λ1 represents a weight of the BCE loss in a total loss for the classification loss, λ2 represents a weight of the DFL loss in the total loss for the localization loss, and λ3 represents a weight of the CIOU loss in the total loss for the another localization loss (see, e.g., sec. 2, which teaches that the loss calculation is a weighted combination of BCE, DFL and CIOU losses); and
S5, detecting the defects in the hydro turbine top cover based on the improved YOLOv8 model (see, e.g., sec. 4.7-4.8, which teach detecting defects using the enhanced model).
While Wang as applied detecting defects on the road, it does not explicitly teach that the defects are in the hydro turbine top cover. Wang 2 in the analogous art teach detecting a fault associated with a top cover of a water turbine using a convolution neural network (see, e.g., lines 23-32 on page 2 of Wang2).
It would have been obvious to use Wang’s method to detect a water turbine fault as taught by Wang 2 because doing so would effectively improve the safety and reliability of the water turbine (see, e.g., lines 8-14 on page 2 of Wang2).
While Wang in view of Wang2 does not explicitly teach, Weinberger in the analogous art teaches that in the step S1, the processing hydro turbine top cover defect images comprises:
cropping the hydro turbine top cover defect images which are taken on-site to obtain cropped images (see, e.g., lines 47-67 in col. 8 and lines 1-10 in col. 9 of Weinberger, which teach synthetically altering, e.g., cropping, images for anomaly detection), and
expanding the cropped images by image enhancement techniques to obtain the hydro turbine top cover defect data set (see, e.g., lines 47-67 in col. 8 and lines 1-10 in col. 9 of Weinberger, which teach synthetically altering, e.g., zooming, images to enhance flaws or defects depicted therein).
It would have been obvious to enhance the defect images as taught by Weinberger because doing so would yield predictable results of generating more realistic images and more accurately trained model (see MPEP 2143(I)(D) and lines 47-67 in col. 8 and lines 1-10 in col. 9 of Weinberger).
While Wang in view of Wang2 and Weinberger does not explicitly teach, Zhang in the analogous art teaches dividing the input dataset into a train set, a validation set and a test set based on a ratio of 6:2:2 (see, e.g., par. 50 of Zhang).
It would have been obvious to divide the input dataset as taught by Zhang because doing so would suitably train the model (see par. 50 of Zhang).
For claim 19, Wang in view of Wang 2, Weinberger and Zhang teaches that the neck network adopts a FPN + PAN structure (see, e.g., sec. 3.1 and FIG. 3 of Wang, which teach improving a FPN + PAN structure).
Claim(s) 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang in view of Wang2, Weinberger and Zhang and further in view of a non-patent literature by H. Wang, G. Yang, X. Hao and L. Geng, entitled "Automotive Gear Defect Detection Method based on Yolov8 Algorithm," published in 2023 Asia Symposium on Image Processing (ASIP), which was held from 06/15/2023 to 06/17/2023 (hereinafter Wang3).
For claim 20, while Wang in view of Wang2, Weinberger and Zhang does not explicitly teach, Wang3 in the analogous art teaches that the CBAM module comprises a CAM and a SAM, the CAM is configured to focus on feature information, and the SAM is configured to focus on target position information (see, e.g., sec. III, A of Wang3, which teaches that the CBAM includes the channel attention module (CAM) module that outputs feature information and the SAM module that adds position information to the feature information).
It would have been obvious to one of ordinary skill in the art to modify Wang1 in view of Wang2 to use the CBAM structure as taught by Wang3 because doing so would improve capturing the key features of defects and to suppress the interference information (see sec. III, A of Wang3).
Allowable Subject Matter
Claims 10-16 are allowed.
In regard to claim 10, when considered as a whole, prior art of record fails to disclose or render obvious, alone or in combination:
“A method for detecting defects in a hydro turbine top cover based on an improved YOLOv8 model, comprising:
…
S3, training a network for detecting the defects in the hydro turbine top cover based on YOLOv8-CBAM to obtain a YOLOv8-CBAM model as the improved YOLOv8 model, specifically comprising:
inputting the hydro turbine top cover defect data set into the network for training, wherein a total loss function of the YOLOv8-CBAM model comprises:
a BCE loss, a DFL loss, and a CIOU loss, and a calculation formula of the total loss function is as follows:
LOSS=λ1LOSSBCE+λ2LOSSDFL+λ3LOSSCIOU
where LOSSBCE represents a classification loss, LOSSDFL represents a localization loss, LOSSCIOU represents another localization loss, λ1 represents a weight of the BCE loss in a total loss for the classification loss, λ2 represents a weight of the DFL loss in the total loss for the localization loss, and λ3 represents a weight of the CIOU loss in the total loss for the another localization loss;
wherein the classification loss LOSSBCE is calculated by using a binary cross-entropy loss function, as follows:
LOSSBCE=-ω[ynlog xn+(1-yn)log(1-xn)]
where ω represents a weight, yn represents a true value of a n-th sample, and xn represents a predicted value of the n-th sample;
wherein the localization loss LOSSDFL is calculated by using a DFL loss function as follows:
LOSSDFL=-(yn+1-xn)log(φn)+(xn-yn)log(φn+1)]
φn=yn+1-xn/yn+1-yn, φn+1=xn-yn/yn+1-yn
where yn+1 represents a true value of a n+1-th sample;
wherein the another localization loss LOSSCIOU is calculated by using a CIOU loss function as follows:
LOSSCIOU=IOU-p2(b,bgt)/c2-αv
α=v(v/1-IOU)
v=4/π2(arctan wgt/hgt-arctan w/h)2
where IOU represents an intersection over union between a true box and a predicted box, b represents a center point position of the predicted box, bgt represents a center point position of the true box, p represents a Euclidean distance between the center points of the predicted box and the true box, c represents a diagonal distance of a smallest bounding box that contains both the predicted box and the true box, a represents a trade-off parameter, v represents a similarity of aspect ratios of the predicted box and the true box, wgt represents a width of the true box, hgt represents a height of the true box, w represents a width of the predicted box, and h represents a height of the predicted box; and
S4, detecting the defects in the hydro turbine top cover based on the improved YOLOv8 model.”
In regard to claims 11-16, they are allowed for their dependencies to claim 10.
Claims 3, 4, 6-9, and 18 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.
In regard to claim 3, when considered as a whole, prior art of record fails to disclose or render obvious, alone or in combination:
“The method as claimed in claim 1, wherein the classification loss in the step S4 is calculated by using a binary cross-entropy loss function, as follows: LOSSBCE= -ω[ynlog xn+(1 - yn)log(1 - xn)] where ω represents a weight, yn represents a true value of a n-th sample, and xn represents a predicted value of the n-th sample.”
In regard to claim 4, when considered as a whole, prior art of record fails to disclose or render obvious, alone or in combination:
“The method as claimed in claim 1, wherein the localization loss LOSSDFL is calculated by using a DFL loss function, the another localization loss LOSSCIOU is calculated by using a CIOU loss function, and the localization loss LOSSDFL and the another localization loss LOSSCIOU are as follows:
LOSSDFL=-(yn+1-xn)log(φn)+(xn-yn)log(φn+1)]
φn=yn+1-xn/yn+1-yn, φn+1=xn-yn/yn+1-yn
LOSSCIOU=IOU-p2(b,bgt)/c2-αv
α=v(v/1-IOU)
v=4/π2(arctan wgt/hgt-arctan w/h)2
where yn represents a true value of a n-th sample, yn+1 represents a true value of a n+1-th sample, xn represents a predicted value of a n-th sample, IOU represents an intersection over union between a true box and a predicted box, b represents a center point position of the predicted box, bgt represents a center point position of the true box, p represents a Euclidean distance between the center points of the predicted box and the true box, c represents a diagonal distance of a smallest bounding box that contains both the predicted box and the true box, α represents a trade-off parameter, v represents a similarity of aspect ratios of the predicted box and the true box, wgt represents a width of the true box, hgt represents a height of the true box, w represents a width of the predicted box, and h represents a height of the predicted box.“
In regard to claims 6-9, each of these claims depends on objected claim 5. Therefore, by virtue of their dependency, claims 6-9 are also indicated as objected subject matter.
In regard to claim 18, when considered as a whole, prior art of record fails to disclose or render obvious, alone or in combination:
“The method as claimed in claim 17, wherein the backbone network is configured to extract features from input images to obtain feature maps, and the backbone network comprises five CBS modules, four C2f modules and a SPFF module;
the neck network is configured to perform multi-scale feature fusion on the feature maps to obtain fused features and send the fused features to the head network, and the neck network comprises five CBS modules, six C2f modules, three UnSample modules, six ConCat module and a CBAM module; and
the head network is configured to predict the input images, and the head network comprises four detection heads; and
wherein each CBS module comprises: a convolution layer, a batch normalization layer connected with the convolution layer, and an activation function layer connected with the batch normalization layer; and each C2f module comprises:
two CBS modules, a spilt layer, three bottleneck layers, and a concatenate layer, one of the two CBS module is connected to the spilt layer, the spilt layer is connected to the three bottleneck layers, and the three bottleneck layers are connected to the concatenate layer, and the concatenate layer is connected to the other CBS module.“
Additional Citations
The following table lists several references that are relevant to the subject matter claimed and disclosed in this Application. The references are not relied on by the Examiner, but are provided to assist the Applicant in responding to this Office action.
Citation
Relevance
Gao et al. (us pat. pub. 2025/0061719)
Describes a method for work zone management. One embodiment includes the steps of providing one or more images of a work zone, analyzing the one or more images to detect one or more work-zone related objects within the work zone, sizing the detected work-zone related objects by comparing the detected objects to known sizes of common work zone equipment to establish a scale, calculating estimated positions of the one or more work-zone related objects, mapping the one or more work-zone related objects to a topological map, calculating a topology complexity score based on the topological map, and determining whether the work zone is an organized work zone or a random accumulation of work zone objects, based on the topology complexity score.
Jarvis et al. (us pat. pub. 2024/0284922)
Describes carcass processing systems, both manually and automatic, in which a mechanism is situated for splitting or removing a portion of the carcass as it is supported on a carcass rail in a carcass processing facility. In one embodiment, the invention relates to a method and apparatus for monitoring the quality of processing suspended carcasses manually or automatically (robotically) as the carcass is moved along a defined path. More specifically, the invention implements a machine vision architecture, incorporating machine learning and/or artificial intelligence (AI) and data analysis to develop a state-of-the-art vision-based quality control and audit system, and method of performing the same, for the purpose of providing consistent information on the quality of product generated from the manual or automatic (robotic) process, such that plant-to-plant and auditor-to-auditor variabilities are minimized.
Li et al. (us pat. pub. 2024/0331190)
Describes a computer-implemented method for automated wellhead monitoring using imaging and computer vision. In one embodiment, the method includes establishing a baseline for a wellhead using image dimension and perspective calibration. The method also includes constructing at least one computer vision model using images or video from a wellhead database and inputting unseen wellhead images to the trained at least one computer vision model. Additionally, the method includes extracting a wellhead shape and geometric information of the wellhead in the unseen wellhead images and estimating wellhead displacement and growth based on the extracted images.
Table 1
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. See Table 1 and form 892.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WOO RHIM whose telephone number is (571)272-6560. The examiner can normally be reached Mon - Fri 9:30 am - 6:00 pm et.
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, Henok Shiferaw can be reached at 571-272-4637. 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.
/WOO C RHIM/Examiner, Art Unit 2676
Prosecution Insights