Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted by applicant dated 09/25/2024 have been considered by the examiner.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim 16 rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because the claim(s) is directed to a signal per se/software per se. Claim 16 recites the limitation “A computer program product loadable in a non-transitory memory” in line 1. There is insufficient antecedent basis for this limitation in the claim. The claim is directed to a "computer program product" without requiring any physical, tangible, or non-transitory computer-readable memory or medium.
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.
Claim 16 is 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.
Claim 16 recites the limitation “the at least one processor”. There is insufficient antecedent basis for this limitation ("the at least one processor"), it is lacks a prior introductory reference ("a processor") in the claim.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-5, 7, 12, 15, 16 are rejected under 35 U.S.C. 103 as being unpatentable over Rajagopalan (US 10834128 B1) in view of Sutton (US 20220027428 A1) in view of Hohwald (US 11017019 B1) in view of YUN (CN 103442014 A).
Note that citations to the Yun reference are made to the English translation of the document included with this Office action.
Regrading claim 1, Rajagopalan teaches a method for detecting phishing attacks, comprising:
- determining whether said source data are included in a first list, wherein said first list includes data of authorized sources (Col 12 lines 22-45. FIG. 6: the URL extractor 103 can pass the received URL through a series of filters at 603 including blacklist filters, whitelist filters, signature filters, and/or other suitable filters. At conditional statement 607 the URL extractor 103 determines whether there was a positive match for the received URL based on the applied filters.);
- in response to a determination that said source data are not included in said first list (Fig. 3. Fig. 6. Col 12 lines 48-58: when none of the filters applied by the URL extractor 103 match the received URL, the URL extractor 103 takes an image of a web element or webpage. The URL extractor 103 then sends the image to the phishing detector 101.):
- generating an image that corresponds to a rendering of said source code (Fig. 2. Col 5 lines 64-67 Col 6 lines 1-12: the image extractor component 209 can produce an image of web elements associated with a URL. For instance, image extractor component 209 can render and/or retrieve web elements based on the URL and take an image or screenshot of the rendered and/or retrieved web elements. The image produced by the image extractor component 209 can include the entire content of a webpage, substantial portion of the content of a webpage, or a portion of a webpage. In some other instances, the image produced by the image extractor component 209 can include images of code included in, for example, script files, libraries or other suitable computer executable code that can be retrieved based on the URL. URL extractor server 103 can further include network interface 203 configured to communicate with phishing detector server 101 and various other compute devices connected to network 105.);
- determining via an object classifier a list of objects in said image (Fig. 2. Col 5 lines 64-67 Col 6 lines 1-12 Col 7 lines 22-32: the image extractor component 209 can produce an image of web elements associated with a URL. For instance, image extractor component 209 can render and/or retrieve web elements based on the URL and take an image or screenshot of the rendered and/or retrieved web elements. The image produced by the image extractor component 209 can include the entire content of a webpage, substantial portion of the content of a webpage, or a portion of a webpage. In some other instances, the image produced by the image extractor component 209 can include images of code included in, for example, script files, libraries or other suitable computer executable code that can be retrieved based on the URL. image analyzer 303 can extract one or more web elements (e.g., logos or logotypes) from an image of a webpage.), wherein said object classifier comprises:
- a feature extractor comprising a convolutional neural network configured to receive said image and provide in output a plurality of feature maps with different sizes (Col 8: a convolution layer performs a linear convolution over an image or feature map via a linear filter; where custom character is an array of real numbers of M×N pixels and K channels. The filter f extracts feature from an input image or a feature map and the resulting array custom character preserves, for example, spatial relationships between feature elements included in custom character. Differently stated, a convolutional layer can be used to convolve an image with a set of features and produce a stack of filtered images or feature maps.), wherein each cell of the feature map comprises a vector of features (Col 8: a convolution layer performs a linear convolution over an image or feature map via a linear filter; where custom character is an array of real numbers of M×N pixels and K channels. The filter f extracts feature from an input image or a feature map and the resulting array custom character preserves, for example, spatial relationships between feature elements included in custom character. Differently stated, a convolutional layer can be used to convolve an image with a set of features and produce a stack of filtered images or feature maps. At the fully connected classifier layer 411, the feature map is converted into a vector.),
- a classifier configured to estimate the presence of an object in each cell of each aggregated feature map as a function of the respective aggregated feature vector (Col 8: At the fully connected classifier layer 411, the feature map is converted into a vector. Each value in the vector is used as a weight to compute similarity scores further used to determine whether a web element (e.g., a webpage, logo, logotype, or logomark) is part of a phishing cyber-attack. In some instances, feature maps produced at convolutional blocks 401-409 and weights produced at the fully connected layer 411 are computed via back propagation and gradient descent techniques.), wherein said classifier is configured to identify a plurality of object classes and provide for each cell and each object class respective data indicating the probability that the respective cell includes an object that belongs to the respective object class, wherein said classifier is trained to detect a plurality of reference logos, wherein each object class corresponds to a respective reference logo (Col 8 lines 63-67 Col 9 lines 1-7: at the fully connected classifier layer 411, the feature map is converted into a vector. Each value in the vector is used as a weight to compute similarity scores further used to determine whether a web element (e.g., a webpage, logo, logotype, or logomark) is part of a phishing cyber-attack. In some instances, feature maps produced at convolutional blocks 401-409 and weights produced at the fully connected layer 411 are computed via back propagation and gradient descent techniques.);
- analyzing said probabilities to verify whether said image includes a main reference logo (Col 8 lines 63-67 Col 9 lines 1-7: At the fully connected classifier layer 411, the feature map is converted into a vector. Each value in the vector is used as a weight to compute similarity scores further used to determine whether a web element (e.g., a webpage, logo, logotype, or logomark) is part of a phishing cyber-attack. In some instances, feature maps produced at convolutional blocks 401-409 and weights produced at the fully connected layer 411 are computed via back propagation and gradient descent techniques.) via the steps of:
- selecting the object class that has the highest probability (Col 9 lines 43-60: The fully connected processing layer 417 outputs a similarity score for each image received from the training set 415. The scores are then processed by phishing classifier 219, which produces a classification value (e.g., malicious, or unknown) based on whether the similarity score computed for an image has reached a predetermined threshold that is associated with a known or learned phishing cyber-attack. In some instances, when the similarity score reaches the predetermined threshold of a known or learned phishing cyber-attack then, the phishing classifier 219 classifies the image as malicious. In some other instances, when the similarity score does not reach a predetermined threshold of a known or learned phishing cyberattack then, the phishing classifier 219 classifies the image as unknown. It is appreciated that an alternative and/or additional implementation can produce similarity scores with respect to images of web elements learned to be legitimate images, that is, learned by CNN 400B to be part of a web campaign of a trusted entity.),
- determining whether the probability of the selected class is greater than a threshold (Col 9 lines 43-60: The fully connected processing layer 417 outputs a similarity score for each image received from the training set 415. The scores are then processed by phishing classifier 219, which produces a classification value (e.g., malicious, or unknown) based on whether the similarity score computed for an image has reached a predetermined threshold that is associated with a known or learned phishing cyber-attack. In some instances, when the similarity score reaches the predetermined threshold of a known or learned phishing cyber-attack then, the phishing classifier 219 classifies the image as malicious. In some other instances, when the similarity score does not reach a predetermined threshold of a known or learned phishing cyberattack then, the phishing classifier 219 classifies the image as unknown. It is appreciated that an alternative and/or additional implementation can produce similarity scores with respect to images of web elements learned to be legitimate images, that is, learned by CNN 400B to be part of a web campaign of a trusted entity.),
- in response to a determination that the probability of the selected class is greater than said threshold, selecting the selected object class as the main reference logo (Col 9 lines 43-60, Col 13 lines 6-19: The scores are then processed by phishing classifier 219, which produces a classification value (e.g., malicious, or unknown) based on whether the similarity score computed for an image has reached a predetermined threshold that is associated with a known or learned phishing cyber-attack. the object analyzer 305 predicts, at 623, a brand associated with logos or logotypes identified on the received screenshot based on an object detection process executed via a CNN as described with respect to FIG. 4B. In some instances, when the object analyzer 305 predicts that the identified logo or logotypes are associated with a brand or pattern known to be characteristic of a phishing cyber-attack (e.g., a positive match with a web element known to be malicious), at 625, the logo or logotypes, the screenshot, and the received URL are classified as malicious; otherwise, they are classified as unknown. As previously discussed, then the flow can continue from 627 or 629 to produce and send an alert to, for example, a user or users who received URL at 621.), and
- in response to a determination that the probability of the selected class is not greater than said threshold, select no main reference logo (Col 13 lines 6-19: the object analyzer 305 predicts, at 623, a brand associated with logos or logotypes identified on the received screenshot based on an object detection process executed via a CNN as described with respect to FIG. 4B. In some instances, when the object analyzer 305 predicts that the identified logo or logotypes are associated with a brand or pattern known to be characteristic of a phishing cyber-attack (e.g., a positive match with a web element known to be malicious), at 625, the logo or logotypes, the screenshot, and the received URL are classified as malicious; otherwise, they are classified as unknown. As previously discussed, then the flow can continue from 627 or 629 to produce and send an alert to, for example, a user or users who received URL at 621.);
- in response to said selecting the selected object class as main reference logo, classifying said web page or said email as phishing attack (Col 13 lines 6-19: The object analyzer 305 predicts, at 623, a brand associated with logos or logotypes identified on the received screenshot based on an object detection process executed via a CNN as described with respect to FIG. 4B. In some instances, when the object analyzer 305 predicts that the identified logo or logotypes are associated with a brand or pattern known to be characteristic of a phishing cyber-attack (e.g., a positive match with a web element known to be malicious), at 625, the logo or logotypes, the screenshot, and the received URL are classified as malicious; otherwise, they are classified as unknown. As previously discussed, then the flow can continue from 627 or 629 to produce and send an alert to, for example, a user or users who received URL at 621.).
Rajagopalan does not disclose receiving a source code of a web page or email, and respective source data identifying a sender of the web page or email, respectively.
Sutton teaches receiving a source code of a web page or email, and respective source data identifying a sender of the web page or email (Col 12 lines 22-26. FIG. 6: At 601, URL extractor 103 receives a URL from a source compute device, for instance, a URL included in an email. In some implementations, the URL extractor 103 can pass the received URL.), respectively.
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan with the teachings of Hohwald to include receiving a source code of a web page or email, and respective source data identifying a sender of the web page or email, respectively in order to enable the model to ensure that a message originates from a legitimate source.
Rajagopalan in view of Sutton does not disclose a feature aggregator configured to generate one or more aggregated feature maps with different sizes as a function of said feature maps, wherein each cell of a respective aggregated feature map comprises a vector of aggregated features.
Hohwald does teaches a feature aggregator configured to generate one or more aggregated feature maps with different sizes as a function of said feature maps, wherein each cell of a respective aggregated feature map comprises a vector of aggregated features (Col 10 lines 4-43: the process 300 begins by proceeding from start step to step 301 when a set of training data 248 (e.g., training images) is fed through a convolutional neural network 240. For example, the convolutional neural network 240 can consist of a stack of eight layers with weights, the first five layers being convolutional layers and the remaining three layers being fully-connected layers. The set of training data 248 can be fixed-size 242×242 pixel Black-White image data or Red-Green-Blue (RGB) image data. In one or more implementations, the set of training data 248 includes a data file containing pixel data for each training image. The set of training data 248 may include a different set of training images for each style class. For example, the set of training data 248 may include a first set of training images representing a candid style class, and a second set of training images representing a vector art style class. The number of sets (or instances) of the training data included in the set of training data 248 may be an arbitrary number and may vary depending on implementation. Subsequently, in step 302, the convolutional neural network 240 transforms pixel data of each training image in the set of training images into a feature descriptor vector. The convolutional neural network 240 extracts feature descriptors from the training images. The convolutional neural network 240 processes the set of training data 248 in order to learn to identify a correlation between an image and a style classification by analyzing pixel data of the image. The extracted features (or feature descriptor vectors) may be then fed into a multinomial logistic regression to map them to respective image style classes.).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton with the teachings of Hohwald to include a feature aggregator configured to generate one or more aggregated feature maps with different sizes as a function of said feature maps, wherein each cell of a respective aggregated feature map comprises a vector of aggregated features in order to enable the model to capture, resize, and combine complex, multi-level features from different layouts.
Rajagopalan in view of Sutton in view of Hohwald does not disclose in response to said selecting no main reference logo, classifying said web page or said email as legitimate.
YUN teaches in response to said selecting no main reference logo, classifying said web page or said email as legitimate (Para [0089]- [0090]: Step 4: Match the webpage content corresponding to the current website address with the exclusive logo images embedded in all whitelisted websites, calculate the similarity, and obtain the maximum similarity value; Step 5: Determine if the maximum similarity value is greater than the preset threshold. If it is, determine that the current website address is a suspected counterfeit website and add a suspected counterfeit mark to the website address; otherwise, determine that the current website address is not a suspected counterfeit website and add a non-suspected counterfeit mark.).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton in view of Hohwald with the teachings of YUN to include in response to said selecting no main reference logo, classifying said web page or said email as legitimate in order to distinguish trusted website from malicious threats.
Regarding claim 2, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 1, comprising:
- determining whether said source data are included in a second list, wherein said second list includes data of malicious sources (Rajagopalan Col 12 lines 22-45. FIG. 6: the URL extractor 103 can pass the received URL through a series of filters at 603 including blacklist filters, whitelist filters, signature filters, and/or other suitable filters. At conditional statement 607 the URL extractor 103 determines whether there was a positive match for the received URL based on the applied filters.); and
- in response to determining that said source data are included in said second list, classifying said web page or said email as a phishing attack (Rajagopalan Col 12 lines 22-45. FIG. 6: the URL extractor 103 can pass the received URL through a series of filters at 603 including blacklist filters, whitelist filters, signature filters, and/or other suitable filters. At conditional statement 607 the URL extractor 103 determines whether there was a positive match for the received URL based on the applied filters. For example, if the URL is associated with a phishing cyber-attacker, in such a case the URL is classified as malicious at 605.).
Regarding claim 3, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 1, comprising:
- in response to said selecting the selected object class as main reference logo, determining whether said source data are included in a third list, wherein said third list includes for said selected object class authorized source data (Rajagopalan Claim 1);
- in response to a determination that said source data are included in said third list, classifying said web page or said email as legitimate (Rajagopalan Claim 1); and
- in response to a determination that said source data are not included in said third list, classifying said web page or said email as a phishing attack (Rajagopalan Claim 1).
Regarding claim 4, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 1, wherein said receiving said source code and said source data comprises:
- receiving said source code and said source data of a web page via the HTTP protocol; and/or
- receiving said source code and said source data of an email via the POP3, IMAP or SMTP protocol (Rajagopalan Col 5 lines 3-15: URL extractor 103 can inspect emails sent to an email account associated with user 115 and extract URLs included in such an email at a Simple Mail Transfer Protocol (SMTP) mail server).
Regarding claim 5, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 1, wherein said receiving said source code and said source data comprises:
- intercepting an HTTP communication and extracting said source code and said source data a web page from said HTTP communication; and/or
- intercepting a POP3, IMAP or SMTP communication and extracting said source code and said source data of an email from said POP3, IMAP or SMTP communication (Rajagopalan Col 5 lines 3-15: URL extractor 103 can inspect emails sent to an email account associated with user 115 and extract URLs included in such an email at a Simple Mail Transfer Protocol (SMTP) mail server).
Regarding claim 7, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 1, wherein said classifier comprises:
- for each aggregated feature map, a respective sub-classifier configured to receive the aggregated feature vector a cell of the respective aggregated feature map and provide in output for each object class respective data indicating a respective probability that the cell comprises an object belonging to the respective object class (Rajagopalan Col 8 lines 63-67 Col 9 lines 1-7: At the fully connected classifier layer 411, the feature map is converted into a vector. Each value in the vector is used as a weight to compute similarity scores further used to determine whether a web element (e.g., a webpage, logo, logotype, or logomark) is part of a phishing cyber-attack.).
Regarding claim 12, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to Claim 1, the method further comprising:
- analyzing said source code of a web page or email to detect links (Rajagopalan Col 12 lines 22-26. FIG. 6: At 601, URL extractor 103 receives a URL from a source compute device, for instance, a URL included in an email. In some implementations, the URL extractor 103 can pass the received URL.), and repeating the following steps for each detected link:
- requesting the source code for the respective link (Rajagopalan Col 12 lines 22-26. FIG. 6: At 601, URL extractor 103 receives a URL from a source compute device, for instance, a URL included in an email. In some implementations, the URL extractor 103 can pass the received URL.), and
- repeating the steps of the method of Claim 1 for each requested source code (Rajagopalan Col 12 lines 22-26. FIG. 6: At 601, URL extractor 103 receives a URL from a source compute device, for instance, a URL included in an email. In some implementations, the URL extractor 103 can pass the received URL.).
As per claim 15, the claim claiming a system corresponding to the method
claim 1 above, and they are rejected, at least for the same reasons.
As per claim 16, the claim claiming a computer program product corresponding to the method
claim 1 above, and they are rejected, at least for the same reasons.
Claim(s) 6, 8, 10 are rejected under 35 U.S.C. 103 as being unpatentable over Rajagopalan (US 10834128 B1) in view of Sutton (US 20220027428 A1) in view of Hohwald (US 11017019 B1) in view of YUN (CN 103442014 A) in view of YU (CN 110136198 A).
Regarding claim 6, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to any of the previous Claim 1.
Rajagopalan in view of Sutton in view of Hohwald in view of YUN does not explicitly disclose wherein said classifier (1304) is implemented with YOLOv5 or YOLOv7.
YU teaches wherein said classifier (1304) is implemented with YOLOv5 or YOLOv7 (Para [0203))
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton in view of Hohwald in view of YUN with the teachings of YU to include wherein said classifier (1304) is implemented with YOLOv5 or YOLOv7 in order to improving both speed and accuracy in detecting objects.
Regarding claim 8, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 7.
Rajagopalan in view of Sutton in view of Hohwald in view of YUN does not explicitly disclose wherein said sub-classifier is configured to provide data indicating a relative position of an object with respect to the respective cell, and data indicating a relative size of an object with respect to the respective cell.
YU teaches wherein said sub-classifier is configured to provide data indicating a relative position of an object with respect to the respective cell, and data indicating a relative size of an object with respect to the respective cell (Para [0124]: the size of the image region corresponding to the target object is obtained; the model parameters of the classifier model are set according to the size of the image region corresponding to the target object; wherein the model parameters include at least one of the following: the size of the convolution kernel, the stride of the convolution kernel, and the number of convolution kernel.).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton in view of Hohwald in view of YUN with the teachings of YU to include said sub-classifier is configured to provide data indicating a relative position of an object with respect to the respective cell, and data indicating a relative size of an object with respect to the respective cell in order to improves detection accuracy of different sizes, and increases computational efficiency by focusing on relevant features.
Regarding claim 10, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 7.
Rajagopalan in view of Sutton in view of Hohwald in view of YUN does not explicitly disclose wherein said sub-classifier is configured to provide said data indicating a relative position of an object with respect to the respective cell and said data indicating a relative size of an object with respect to the respective cell only during a training phase.
YU teaches wherein said sub-classifier is configured to provide said data (Para [0124]: the size of the image region corresponding to the target object is obtained; the model parameters of the classifier model are set according to the size of the image region corresponding to the target object; wherein the model parameters include at least one of the following: the size of the convolution kernel, the stride of the convolution kernel, and the number of convolution kernel.).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton in view of Hohwald in view of YUN with the teachings of YU to include wherein said sub-classifier is configured to provide said data .
Claim(s) 9 is rejected under 35 U.S.C. 103 as being unpatentable over Rajagopalan (US 10834128 B1) in view of Sutton (US 20220027428 A1) in view of Hohwald (US 11017019 B1) in view of YUN (CN 103442014 A) in view of YU (CN 110136198 A) in view of Chen (US 20210026355 A1).
Regarding claim 9, Rajagopalan in view of Sutton in view of Hohwald in view of YUN in view of YU teaches the method according to Claim 8, wherein said object classifier
- a post-processing module (Rajagopalan Col 11 lines 60-67 Col 12 lines 1-21: web elements collected as images can be clustered at 503 using, for instance, one or more visual similarity technique. Visual similarity techniques are a type of content-based image analysis that involves computing similarity measures between images based on features of such images. Examples of image features often used in visual similarity techniques include form of edges included in an image, colors, types of textures determined based on pixel values, shape measures, salient point or regions and other suitable features. Thereafter, a clustering process can be executed to group images based on the computed distances into clusters or groups. Thereafter, or in parallel, each retrieved image can be labeled using, for example, an optical character recognition (OCR) process at 505. The OCR process executed at 505 produces one or more labels based on the content of each retrieved image. OCR is performed at 505 only on the centroid of each cluster from a set of clusters of images and label all the images in that cluster with a label extracted from the centroid. In some other implementations, one or more of the image analyzers 303 and/or the object analyzer 305 can be trained and used to performed the automatic labeling process shown at 505);
- data identifying the confidence of the estimate that said image comprises the respective object (Rajagopalan Col 9 lines 43-60).
Rajagopalan in view of Sutton in view of Hohwald in view of YUN in view of YU does not explicitly disclose wherein said post-processing module is configured to apply a non-maximum suppression to the data provided by said classifier: data identifying an object class selected from said object classes; optionally data indicating the position and size of the object in said image.
Chen teaches wherein said post-processing module is configured to apply a non-maximum suppression to the data provided by said classifier (Para [0053]: the machine learning model(s) 108 may predict multi-channel classification data (e.g., the class confidence data 110, the instance confidence data 112), multi-channel object instance data (e.g., the instance regression data 111), and/or a depth map (e.g., the depth data 113) from a particular input (e.g., the input data 106). Some possible training techniques are described in more detail below. In operation, the outputs of the machine learning model(s) 108 may be decoded (e.g., via post-processing 114) to identify bounding shapes identifying the locations, geometry, and/or orientations of detected objects, class labels for detected objects, instance labels for detected objects, and/or range to detected objects. In some embodiments, since object instance data may be noisy and/or may produce multiple candidates, bounding shapes may be generated using non-maximum suppression, density-based spatial clustering of application with noise (DBSCAN), and/or another function.):
- data identifying an object class selected from said object classes (Para [0053]: In operation, the outputs of the machine learning model(s) 108 may be decoded (e.g., via post-processing 114) to identify bounding shapes identifying the locations, geometry, and/or orientations of detected objects, class labels for detected objects, instance labels for detected objects, and/or range to detected objects. In some embodiments, since object instance data may be noisy and/or may produce multiple candidates, bounding shapes may be generated using non-maximum suppression, density-based spatial clustering of application with noise (DBSCAN), and/or another function.); and
- optionally data indicating the position and size of the object in said image (Para [0053]: In operation, the outputs of the machine learning model(s) 108 may be decoded (e.g., via post-processing 114) to identify bounding shapes identifying the locations, geometry, and/or orientations of detected objects, class labels for detected objects, instance labels for detected objects, and/or range to detected objects. In some embodiments, since object instance data may be noisy and/or may produce multiple candidates, bounding shapes may be generated using non-maximum suppression, density-based spatial clustering of application with noise (DBSCAN), and/or another function.).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton in view of Hohwald in view of YUN in view of YU with the teachings of Chen to include wherein said post-processing module is configured to apply a non-maximum suppression to the data provided by said classifier: data identifying an object class selected from said object classes; optionally data indicating the position and size of the object in said image in order to maintain only the most accurate prediction for each object.
Claim(s) 11 is rejected under 35 U.S.C. 103 as being unpatentable over Rajagopalan (US 10834128 B1) in view of Sutton (US 20220027428 A1) in view of Hohwald (US 11017019 B1) in view of YUN (CN 103442014 A) in view of TAJI (JP 2018147431 A).
Regarding claim 11, Rajagopalan in view of Sutton in view of Hohwald in view of YUN teaches the method according to claim 1.
Rajagopalan in view of Sutton in view of Hohwald in view of YUN does not disclose wherein the height and width of each aggregated feature map
TAJI teaches wherein the height and width of each aggregated feature map (Para [0017]. Para [0034]: among the selectable feature maps, for example, a 28x28 pixel feature map.).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Rajagopalan in view of Sutton in view of Hohwald in view of YUN with the teachings of TAJI to include wherein the height and width of each aggregated feature map is smaller than 50 in order to improve the accuracy of object identification by using lower-resolution feature maps (TAJI Para [0017]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JUDY BAZNA whose telephone number is (703)756-1258. The examiner can normally be reached Monday - Friday 08:30 AM-05:00 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.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Farid Homayounmehr can be reached at (571) 272-3739. 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.
/JUDY BAZNA/Examiner, Art Unit 2495
/FARID HOMAYOUNMEHR/Supervisory Patent Examiner, Art Unit 2495
Prosecution Insights