DETAILED ACTION
Claim(s) 24,25,26 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of Bilenko et al. (US 2013/0204809 A1):
Claim(s) 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1) as applied in claims 24,25,26:
Claim(s) 5 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES):
Claim(s) 6 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES) as applied in claims 5 and 17 further in view of YIN (CN 108133476 B) with Google Patents machine translation:
Claim(s) 9 and 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of Gasper et al. (US 2022/0327119 A1):
Claim(s) 10 and 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of Rogers et al. (US 2021/0117859 A1):
Response to Amendment
The amendment was received 1/26/2026. Claims pending 1-26 pending.
Priority
Applicant’s claim for the benefit of a prior-filed application under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, 365(c), or 386(c) is acknowledged. Applicant has not complied with one or more conditions for receiving the benefit of an earlier filing date under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, 365(c), or 386(c) as follows:
The later-filed application must be an application for a patent for an invention which is also disclosed in the prior application (the parent or original nonprovisional application or provisional application). The disclosure of the invention in the parent application and in the later-filed application must be sufficient to comply with the requirements of 35 U.S.C. 112(a) or the first paragraph of pre-AIA 35 U.S.C. 112, except for the best mode requirement. See Transco Products, Inc. v. Performance Contracting, Inc., 38 F.3d 551, 32 USPQ2d 1077 (Fed. Cir. 1994).
The disclosure of the prior-filed application, Application No. 18/145,301 12/22/2022, fails to provide adequate support or enablement in the manner provided by 35 U.S.C. 112(a) or pre-AIA 35 U.S.C. 112, first paragraph for one or more claims of this application.
Application No. 18/145,301 12/22/2022 does not disclose claim 1’s:
“search configuration parameter”
“custom”:
PNG
media_image1.png
669
825
media_image1.png
Greyscale
PNG
media_image2.png
1097
842
media_image2.png
Greyscale
PNG
media_image3.png
957
741
media_image3.png
Greyscale
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified (detection) function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts (plural) described in the specification [0066] and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified (detection) function without the recital of structure, material, or acts (plural) in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification [0066] and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts (plural) for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts (plural) to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites (detection) function without reciting sufficient (computer) structure, material or acts (plural) to entirely perform the recited (detection) function.
This application includes one or more claim limitations that use the word “means” or “step” but are nonetheless not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph because the claim limitation(s) (prepositionally or adjectivally) recite(s) sufficient structure (a “computer” by definition (Dictionary.com) of claim 1’s “output”1 & “machine learning”2, each comprising “computer”), materials, or acts to entirely perform the recited (underlying) function, y(x). Such claim limitation(s) is/are:
--the custom model is configured to detect3 custom objects--
in claim 4;
--a text-to-visual content model trained to detect4 custom objects corresponding to the custom object label and to detect5 at least one predetermined object for which the at least one teacher model is trained to detect6--
in claim 5;
--the custom model is configured to detect custom objects--
in claim 16;
--the custom model is a text-to-visual content model trained to detect custom objects corresponding to the custom object label and to detect at least one predetermined object for which the at least one teacher model is trained to detect--
in claim 17.
Because this/these claim limitation(s) is/are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are not being interpreted to cover only the corresponding structure, material, or acts described in the specification as performing the claimed function, and equivalents thereof.
If applicant intends to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to remove the (“computer”) structure, materials, or acts that performs the claimed (underlying) function; or (2) present a sufficient showing that the claim limitation(s) does/do not recite sufficient (computer) structure, materials, or acts to perform the claimed (underlying) function (y(x)).
Claim - 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.
Claims 1-26 not rejected under 35 U.S.C. 101 because the claimed invention is directed to improving machine learning technology not without significantly more (streamlined analysis).
Response to Arguments
Claim rejections – 35 USC 102
Applicant's arguments filed 8/12/2025 have been fully considered but they are not persuasive:
Applicant's arguments filed 1/26/2026 have been fully considered but they are not persuasive:
In response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e.,
“having a set of potential features”, pg. 11, 2nd para) are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993):
PNG
media_image4.png
609
822
media_image4.png
Greyscale
Applicants state in page 11, last S that Wei does not teach a model using a smaller domain7 in original claim 24. The examiner respectfully disagrees since Wei (US 11,445,168) teaches a “model” “146” using a classed (via fig. 2:144: “PER DEFECT CATEGORY”) “area”8-“subset”9, understood to be a “categorized”10 domain smaller than the corresponding larger area-set (“set of…areas”):
PNG
media_image5.png
1525
969
media_image5.png
Greyscale
Applicant’s arguments, see remarks, pgs. 12,13 “potential features” filed 1/26/2026, with respect to the rejection(s) of claim(s) 24 under 35 USC 102 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of 35 USC 103:
Claim(s) 24,25,26 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of Bilenko et al. (US 2013/0204809 A1), wherein Bilenko teaches “potential feature” improving Wei’s model (fig. 2:142: “TRAINED MODEL(S)”) and the classified-domain-area- subset-inputting (fig. 2:148A: “LOCAL DEFECTS”-arrows) machine-learning model (fig. 2:146: Machine Learning Model “DEFECT DECTOR(S)”):
PNG
media_image6.png
781
1097
media_image6.png
Greyscale
In response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e., (potential) “features…are a subset”, pg. 12, penult para, 1st S) are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993):
In contrast claim 24 states: “the domain…is a set”.
Note that limitations from applicant’s disclosure are not read in claim 24:
--In11 particular, the domain (i.e.12, the universe of potential features)…--[0056]: meaning:
~the particular domain (in13 other words the particular universe of particular potential particular features)…is a subset~ is not read into claim 24:
PNG
media_image4.png
609
822
media_image4.png
Greyscale
In response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e., “features…as being a subset”, pg. 13, 1st para, 1st S) are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993):
PNG
media_image4.png
609
822
media_image4.png
Greyscale
Claim rejections – 35 USC 103
Claims 1,2,3,4,7,8,11,12,13,14,15,16,19,20 and 23
Applicant’s arguments, see remarks, page 14: regarding “potential features”, filed 1/26/2026, with respect to the rejection(s) in the Office action of 10/23/2025 of claim(s) 1,2,3,4,5,6,11 and 12 and 13,14,15,16,19,20,23 under 35 USC 103 have been fully considered and are persuasive:
Applicants state that Wei cannot teach “a domain used by the custom model is a subset of a domain used by the advanced machine learning model”, that appears directed to original claim 1; however, this statement is in context of “potential features” which is considered persuasive. Therefore, the rejection has been withdrawn.
However, upon further consideration, a new ground(s) of rejection is made in view of:
Claim(s) 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1) as applied in claims 24,25,26.
Claims 5 and 17
Claims 5 and 17 are rejected via said Bilenko:
Claim(s) 5 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES):
Claim 6 and 18
Claims 6 and 18 are rejected via said Bilenko:
Claim(s) 6 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES) as applied in claims 5 and 17 further in view of YIN (CN 108133476 B) with Google Patents machine translation:
Claims 9 and 21
Claims 6 and 18 are rejected via said Bilenko:
Claim(s) 9 and 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of Gasper et al. (US 2022/0327119 A1):
Claims 10 and 22
Claims 6 and 18 are rejected via said Bilenko:
Claim(s) 10 and 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of Rogers et al. (US 2021/0117859 A1):
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) 24,25,26 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of Bilenko et al. (US 2013/0204809 A1):
PNG
media_image7.png
930
300
media_image7.png
Greyscale
Re 24. (Currently Amended), Wei teaches A method for visual content processing, comprising:
applying a custom machine learning model (via fig. 3:308: “CUSTOM MODEL SYSTEM”) to a set of first media (player: fig. 1:120: “MEDIA PLAYER”) content (“delivery system 100”, c.4,ll. 50-55), wherein the custom machine learning model (via fig. 3:308: “CUSTOM MODEL SYSTEM”) is trained (via fig. 1:142: “TRAINED MODEL(S)”) using a set of second media content (“files 320”, c.9,ll. 25-30: fig. 3: 320: “TRAINING VIDEO FILE(S)”), wherein the set of second (training) media content (320) is generated (as “a proper subset of candidate defect frames”, c,11,ll. 10-15) by labeling a plurality of training candidates (or a to-be-inferred defect forming or “comprising at least one labeled defect”, c.11,ll 5-10) based on a plurality of predictions (i.e., machine learning, ML, inferences identifying said subset) output by a at least one teacher (“training by training service 332”14 “candidate”, c. 10,ll. 30-35: fig. 3: “MODEL TRAINING SERVICE”, designed to teach) model (fig. 1:142: candidate-teacher “TRAINED MODEL(S)”), wherein the plurality of (ML) prediction (defect) labels includes at least one instance of a custom (video) object (“custom model” “data objects (such as video files 318)”,c.9,ll. 23-25 & c.7,ll. 25-30) label (or “labeled” “Data 318”, c.9,ll. 5,6, “for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”), wherein the outputs of the custom machine learning model include at least one prediction for the set of first media content (as understood given ML’s “inference results 360B”, c.10,ll.40-45L fig. 3:360B: “INFERENCE RESP.”);
selecting (or extracting15) a (“features 140”, c.3,ll.60-65: fig. 2:140: “FEATURE(S) (E.G., PER FRAME)”) subset of the first media (player) content (fig. 4:408: “RECEIVING AN INFERENCE REQUEST FOR AN INPUT VIDEO FILE”) based on the at least one (inference-training) prediction for the set of first media (player) content output by the (training-inference) custom machine learning model (resulting in fig. 1:142: “TRAINED MODEL(S)”); and
providing (by preference) the (extraction) selected (“features 140”, c.3,ll.60-65: fig. 2:140: “FEATURE(S) (E.G., PER FRAME)”) subset of the first media content as inputs to an advanced machine learning model (fig. 2:146:“DEFECT DETECTOR(S)”), wherein a (“local”-scope, c.5,ll. 55-60: fig. 2:200-210) domain16 used by the custom machine learning model (fig. 2:142: “TRAINED MODEL(S)”) is smaller than a (global-scope) domain (fig. 2:148B: “GLOBAL DEFECTS”) used by the advanced machine learning model (fig. 2:146: deep-ML-“DEFECT DETECTOR(S)”), wherein the (“locations 140A…ID 150B”, c. 6, ll. 35-40/local/global/area: fig. 1: 108: “COMPRESSED (E.G., RESOLUTION FRAMES”) domain of17 each of18 the custom model and of19 the advanced machine learning model is (“directed to”, c. 2,ll, 3rd S: via the arrows in fig. 1) a (sub-)set (fig. 1:144: “CANDIDATE DEFECT FRANE(S)”) of20 potential features21 (“140”, c.4, 1st S: fig. 1:140: “FEATUIRE(S)”) used by the respective model, wherein the set of22 potential features of23 the domain used by the custom model is a subset (144) of24 the set of potential features used by the advanced machine learning model.
Wei does not teach the difference25 of claim 24 of:
potential (features)26.
Bilenko teaches a similar problem of applicants’:
[0003] For example, many domains, such as web search and advertising, utilize sophisticated, computationally expensive learning algorithms and very large labeled datasets, imposing experimentation latency that is a barrier to rapid feature design. Thus, the traditional approaches that re-run the learning algorithm on the labeled data augmented by the potential feature can be computationally costly and time consuming. According to another example, industrial implementations of learning algorithms are typically components within large infrastructure pipelines, which can require significant domain expertise to run. Following this example, potential feature contributors lacking such expertise can be deterred from evaluating their features (e.g., features developed for a different application in the same organization) due to the complexity of adding the potential feature to the training pipeline (e.g., due to logistical costs). Pursuant to yet another example, in some domains, such as medical or marketing applications, potential feature values may be unavailable for the complete training set or may carry non-negligible costs, encouraging evaluation of feature relevance on a data subset before committing to obtaining values of the potential feature for all data (e.g., due to monetary costs).
and the difference (b)) of claim 24 of:
(“Pursuant to yet another example, in some domains, such as medical or marketing applications, potential feature values may be unavailable for the complete training set or may carry non-negligible costs, encouraging evaluation of feature relevance on a data subset before committing to obtaining values of the”) potential (features)27 (“for all data (e.g., due to monetary costs)” [0003] last S: fig. 3:304: “Potential Feature d+1”).
Since Wei teaches a feature, one of skill in the art of features can make Wei’s be as Bilenko’s seeing in the change an “improvement of a previously trained predictor”, Bilenko [0017] 2nd S:
PNG
media_image8.png
1995
1093
media_image8.png
Greyscale
Re 25. (Original), Wei of the combination of Wei,Bilenko teaches The method of claim 24, further comprising:
selecting the plurality of training (frame) candidates from among a set of third media content (comprised by video “frame 8”, c.4,ll.5-10) based on at least one search (“hyper”, c.9,ll. 15-20)parameter, wherein the at least one search (hyper)parameter defines criteria (comprised by said hyperparameter) for identifying the set of third(-frame) media content and for selecting samples (“for (e.g., automated) video quality monitoring”, c.1, 60-65) from among the third set of media content to be used as the training candidates, wherein the set of third media content includes a plurality of portions of media content showing a custom object (“such as video files 318”, c.7,ll. 25-30: fig. 3:318: “VIDEO FILE(S) (E.G., AND OTHER DATA)) corresponding to the custom object label (“for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”).
Re 26. (Original), Wei of the combination of Wei,Bilenko teaches The method of claim 24, wherein each of the (categorization) custom machine learning model (fig. 1:142) and the (categorization) advanced machine learning model (fig. 1:146) is a classifier (via “Utilizing this categorization”, c.2, ll.40-45), wherein the custom machine learning model (142) is configured to output a plurality of first (“categorized…local defects”, c.2,ll.30-35) classes, wherein the advanced machine learning model (146) is configured to output a plurality of second (global) classes, wherein the plurality of first (local) classes is a subset (via a subset of candidates models and a subset of frames: fig. 2:142,144) of the plurality of second (global) classes, wherein each of the plurality of first (local) classes and the plurality of second (global) classes includes a custom object class corresponding to the custom object label28 (or custom defects labeled via “labeled defects 322) for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”).
Claim(s) 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1) as applied in claims 24,25,26:
PNG
media_image9.png
930
417
media_image9.png
Greyscale
Claim 1 is rejected like claim 24:
Re 1. (Currently Amended), Wei teaches A method for visual content processing, comprising:
applying at least one teacher (“training service”-designed to teach--”candidate”, c. 10, ll.31-35) model (figs.1,2,3:142: TRAINED MODEL) to a set of training (“video file”, c.11,ll.10-15) candidates (or defects in fig. 3:318,320: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”) in order to output (fig. 2:arrows) a plurality of (local and global) instances (“for example”, c.2,ll. 15-20) of a custom object (“defects”, c.9,ll. 1-5) label (fig. 3:322: “LABELED DEFECT(S)”29: “for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”), wherein the set of training (video file) candidates (defects in 318,320) is (preferentially) selected (resulting in being “extracted”30, c.3,ll. 60-65: fig. 2:138: “FEATURE EXTRACTOR”) using a (“machine learning”, c.1,l..65 to c.2,l.2) student31 model (represented in fig. 2 as 138: “FEATURE EXTRACTOR”), wherein the (extractor) student model (138) is configured to select (i.e., extract) the (“sub”, c.1,l.65 to c.2,l. 2)set of training (“video”, c.11, ll. 10-15) candidates from among a set of media content (via fig. 1:134: “CONTENT PROVIDER(S)”) based on at least one (grid-)search configuration (“hyper”, c.9,ll. 15-25)parameter32, wherein the at least one (grid-)search configuration (hyper)parameter (exactly or specifically) defines (measuring) criteria for selecting (“video”, c.1,ll. 60-65) samples to be used as the set of training (defect) candidates;
generating (via “the defect detection system 136”, c.6,ll. 50-55: fig. 1: “DEFECT DETECTION SYSTEM”) a first set (“of manifests 116”, c.4,ll. 50-55: fig. 1:116: “MANIFEST IDENTIFYING VIDEO REPRESENTATION(S)”) of media content (via fig. 1:134: “CONTENT PROVIDER(S)”) by labeling (via fig. 3:322: “LABELED DEFECT(S)”) the set of training candidates based on the plurality of (example-feature-extraction) instances of the custom object label (fig. 3:322: “LABELED DEFECT(S)”: “for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”) output by the at least one teacher model (figs. 1,2,3:142: “TRAINED MODEL(S)”; “TRAINED MODEL”, “TRAINED MODEL(S)”);
creating a custom model (fig. 1:142 via a “customer”, c.8,l. 66 to c.9,l. 6) using the at least one teacher model (fig. 1:142), wherein the (customer) custom model is a machine learning model trained (fig. 1:142: “TRAINED MODEL(S)”) using the first set of media content (via fig. 1:134: “CONTENT PROVIDER(S)”);
obtaining a (“representative”, c.1,l.65 to c.2, l. 2) subset of a second set of media content, wherein the (representative) subset of the second set of media content (via fig. 1:134: “CONTENT PROVIDER(S)”) is selected (i.e., extracted) based on outputs (fig. 2: arrows) of the (customer) custom model as applied to the second set of media content, wherein the (arrow) outputs of the (customer) custom model include at least one (inference) first prediction for the (representative) subset of the second set of media content (via fig. 1:134: “CONTENT PROVIDER(S)”); and
applying an advanced machine learning model (fig. 1:146: “DEFECT DECTECTOR(S)”) to the (representative) obtained subset of the second set of media content, wherein a (“local”, c.5,ll. 50-55) domain (fig. 2:200-210) used by the (customer) custom model (142) is a subset of a (“global”, c.6,ll. 20-25) domain used by the advanced machine learning model (fig. 1:146: “DEFECT DECTECTOR(S)”), wherein the domain of each of the custom model and of the advanced machine learning model is a set of potential features used by the respective model, wherein the set33 of potential features of the domain used by the custom model is a subset of the set of potential features used by the advanced machine learning model.
.
Wei does not teach the difference34 of claim 1 of:
a) creating a custom model using the at least one teacher model; and
b) potential (features)35…potential (features)…potential (features).
Da Silva teaches the difference a) of claim 1:
Re 1., A method for visual content processing, comprising:
applying at least one teacher model (fig. 7:704a: “USE TEACHER MODEL TO SELECT ATTACKS”) to a (“new” [0042] 2nd S) set of (discriminator) training candidates (resulting in a “trained” “discriminator” [0043], 2nd S) in order to output a plurality of instances (via an “example” “cat image” [0026]: fig. 2:208,210:cats) of a (“same” [0025] last S) custom object (“object” “image captures” [0040] last S) label (fig. 1: “DOG”), wherein the (new) set of (discriminator) training candidates is selected (i.e., labeled36) using a (discriminator) student model (fig. 1:102: “STUDENT MODEL”), wherein the (discriminator) student model (102) is configured to select (i.e., label) the (new) set of (discriminator) training candidates from among a set of (“video”37 [0012]) media content based on at least one (“accepted” [0064]) search configuration parameter, wherein the at least one (accepted) search configuration parameter defines (accepted) criteria for selecting (i.e., labeling) samples (fig. 7:704a: “USE TEACHER MODEL TO SELECT ATTACKS”: fig. 1:104: SOURCE IMAGE + 108: PERTURBATIONS as labeled/selected by teacher) to be used (in the remainder of fig. 7) as the (new-adversarial) set of (discriminator) training candidates;
generating (“a new dataset” [0042] 2nd S: fig. 7:706a: “GENERATE ADVERSARIAL IMAGES”) a first (new) set of (video) media content by labeling the (new) set of (discriminator) training candidates (as “ ‘noise’ ” [0043]) based on the plurality of (cat) instances (being “misclassified” [0025] last S) of the (same) custom (dog) object label output (via fig. 1: arrows) by the at least one teacher model (figs. 1:7:703a: “LAYER COPIED FROM TEACHER” ; “USE TEACHER MODEL TO SELECT ATTACKS” being the basis for the segmentation/classification student model and also being a “similar, but distinct, task” [0021], 1st S, of student classification);
a) creating a custom (“accurate” [0021] 2nd S) model (fig. 1:102: “STUDENT MODEL” being the recipient of a customization “to further customize”, [0021] 3rd S, teacher models building the student model) using the at least one teacher model (“to create accurate”, [0021] [[2nd]] 3rd S, student models), wherein the custom (accurate) model is a machine learning model (“which may be referred to herein as a ‘discriminator’ “ [0040]: fig. 7:708a: “TRAIN DISCRIMINATOR”) trained using the first (new) set of (video) media content;
obtaining a subset of a second set of (video) media content, wherein the subset of the second set of (video) media content is selected (i.e., labeled) based on outputs (fig. 3:arrows) of the custom (accurate) model (fig. 3:304: “SEGMENTATION MODEL” that “comprises a student model” [0079]) as applied to the second set of (video) media content, wherein the outputs (fig. 3:arrows) of the custom (accurate) model include at least one first prediction (fig. 4,5: “PREDICTIONS”) for the subset of the second set of (video) media content; and
applying an advanced machine (“deep” [0020]) learning model to the obtained subset of the second set of (video) media content, wherein a (“different”, [0024] last S) domain used by the custom (accurate) model is a subset of a (different) domain used by the advanced machine (deep) learning model.
Since Wei suggests training with tuning or the like, c.10,ll. 26-31, one of skill in the art of training can make Wei’s be as de Silva’s transfer learning (i.e., transforming a teacher model into a student model) predictably recognizing the change creating “accurate models for specific tasks”, de Silva [0021] 2nd S, with remedial corrective action, such as creating accurate transfer learning machine learning models (i.e. discriminators) with the task of accurately categorizing conforming frames by type of defect with remedial corrective action of non-conforming noisy frames if need be:
PNG
media_image10.png
1535
1136
media_image10.png
Greyscale
PNG
media_image11.png
1543
1136
media_image11.png
Greyscale
PNG
media_image12.png
1478
1146
media_image12.png
Greyscale
PNG
media_image13.png
1478
1146
media_image13.png
Greyscale
PNG
media_image14.png
1280
1695
media_image14.png
Greyscale
Wei of The combination of Wei,DeSilva does not teach the last difference b) of claim 1:
b) potential (features)38…potential (features)…potential (features).
Bilenko already teaches/makes obvious difference “b)” in the rejection of claim 24.
Re 2., Wei of the combination of Wei,de Silva,Bilenko teaches The method of claim 1, further comprising:
training (via the above illustrated combination of figs. 3 & 7:708a: “TRAIN DISCRIMINATOR”) the custom (accurate-student) model (Wei’s 142,146 as modified via the combination, illustrated above) using the at least one teacher (accurate-teacher) model (142,146 as modified via the combination), wherein a (local) domain used by the custom (accurate) model (142,146 as modified via the combination) is a subset of a (global) domain used by the at least one teacher (accurate) model (142,146 as modified via the combination); and
sending (via arrows in fig. 3), from a second system (fig. 3:300:” PROVIDER NETWORK”) to a first system (fig. 3:306: “INTERMEDIATE NETWORK(S)”), the trained custom model (142) for deployment as the first machine learning (“hosted”, c.10,ll. 35-40) model (fig. 3:336: “HOSTED MODEL(S)”) at the first system (fig. 3:306: “INTERMEDIATE NETWORK(S)”), wherein the first system (fig. 3:306: “INTERMEDIATE NETWORK(S)”) is remote (via a “ ‘cloud’ provider network”, c.7,ll. 5-10) from the second system (fig. 3:300:” PROVIDER NETWORK”).
Re 3., Wei of the combination of Wei,de Silva,Bilenko teaches The method of claim 2, wherein creating the custom model using the at least one teacher model (i.e., transfer learning), further comprises:
labeling (Wei: fig. 3:322) at least a portion of the plurality of training candidates (fig. 3:318,320: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”) with the custom object (“custom model” “data objects (such as video files 318)”,c.9,ll. 23-25 & c.7,ll. 25-30) label (“for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”) based on the output plurality of instances (“for example”, c.2,ll. 15-20) of the custom object (“custom model” “data objects (such as video files 318)”,c.9,ll. 23-25 & c.7,ll. 25-30) label (“for use in creating a custom model”, c.9,ll. 25-30, via fig. 3:309: “USER”) in order to create labeled media content (fig. 3:318,320,322: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”; “LABELED DEFECTS”); and generating (expressed as arrows in fig. 3) the set (with a “representative”, c.1,l.65 to c.2, l. 2, subset) of second media content (via Wei: fig. 1:134: “CONTENT PROVIDER(S)”) based on at least a portion of the labeled media content (fig. 3:318,320,322: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”; “LABELED DEFECTS”).
Re 4., Wei of the combination of Wei,de Silva,Bilenko teaches The method of claim 3, wherein the custom model (Wei: fig. 1:142 via a “customer”, c.8,l. 66 to c.9,l. 6 as modified by the combination) is configured to detect (via fig. 1:136: “DEFECT DETECTION SYSTEM”) custom objects (“custom model” “data objects (such as video files 318)”,c.9,ll. 23-25 & c.7,ll. 25-30: fig. 2:148,150A: “DEFECT(S)”; “FRAME(S) ID”) corresponding to the custom object label (“for use in creating a custom model”, Wei: c.9,ll. 25-30, via fig. 3:309: “USER”) when applied to (extracted) features of media content.
Re 7., Wei of the combination of Wei,de Silva,Bilenko teaches The method of claim 3, wherein the second set of media content (via Wei: fig. 1:134: “CONTENT PROVIDER(S)”) includes exactly one sample (“for feature generation”, Wei: c.3,ll. 20-30) of previously labeled media content (via fig. 3:318,320,322: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”; “LABELED DEFECTS”).
Re 8., Wei of the combination of Wei,de Silva,Bilenko teaches The method of claim 3, further comprising:
selecting (i.e., labeling) the plurality of training candidates (fig. 3:318,320: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”) from among a third (“sub”-,c.3,l. 66 to c.4,l.5: I count four sets in total)set of media content (fig. 1:134: “CONTENT PROVIDER(S)”) based on at least one (“grid”, c.9,ll.15-20) search (hyper)parameter, wherein the at least one (grid-)search (hyper) parameter defines criteria for identifying the third (sub-)set (of four sets) of media content (fig. 1:134: “CONTENT PROVIDER(S)”) and for selecting (i.e., labeling) samples from among the third (sub-)set of media content (fig. 1:134: “CONTENT PROVIDER(S)”) to be used as the training candidates (fig. 3:318,320: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”) wherein the (sub-)set of third(-set) media content (fig. 1:134: “CONTENT PROVIDER(S)”) includes a plurality of (“segment”, c.4,ll. 60-65) portions of media content (fig. 1:134: “CONTENT PROVIDER(S)”) showing a (modeling) custom (video) object (i.e., fig. 3: 318: “VIDEO FILE(S) (E.G., AND OTHER DATA)”) see corresponding rejection of claim 24) corresponding to the custom object label (“for use in creating a custom model”, Wei: c.9,ll. 25-30, via fig. 3:309: “USER”).
Claim 11 is rejected similar to claim 26:
11. The method of claim 1, wherein each of the first machine learning model and a second machine learning model is a classifier, wherein the first machine learning model is configured to output a plurality of first classes, wherein the second machine learning model is configured to output a plurality of second classes, wherein the plurality of first classes is a subset of the plurality of second classes, wherein each of the plurality of first classes and the plurality of second classes includes a custom object class corresponding to the custom object label.
Claim 12 is rejected similar to claim 1:
Re 12. (Currently Amened), Wei of the combination of Wei, DaSilva, Bilenko teaches A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to execute a process, the process comprising:
applying at least one teacher model to a set of training candidates in order to output a plurality of instances of a custom object label, wherein the set of training candidates is selected using a student model, wherein the student model is configured to select the set of training candidates from among a set of media content based on at least one search configuration parameter, wherein the at least one search configuration parameter defines criteria for selecting samples to be used as the set of training candidates;
generating a first set of media content by labeling the set of training candidates based on the plurality of instances of the custom object label output by the at least one teacher model;
creating a custom model using the at least one teacher model, wherein the custom model is a machine learning model trained using the first set of media content;
obtaining a subset39 of a second set of media content, wherein the subset40 of the second set of media content is selected based on outputs of the custom model as applied to the second set of media content, wherein the outputs of the custom model include at least one first prediction for the subset41 of the second set of media content; and
applying an advanced machine learning model to the obtained subset42 of the second set of media content, wherein a domain used by the custom model is a subset43 of a domain used by the advanced machine learning model, wherein the domain of each of the custom model and of the advanced machine learning model is a set of potential features used by the respective model, wherein the set of potential features of the domain used by the custom model is a subset44 of the set of potential features used by the advanced machine learning model.
Claim 13 is rejected similar to claim 1:
Re 13. (Currently Amended) Wei of the combination of Wei, DaSilva, Bilenko teaches A system for visual content processing, comprising:
a processing circuitry; and
a memory, the memory containing instructions that, when executed by the processing circuitry, configure the system to:
apply at least one teacher model to a set of training candidates in order to output a plurality of instances of a custom object label, wherein the set of training candidates is selected using a student model, wherein the student model is configured to select the set of training candidates from among a set of media content based on at least one search configuration parameter, wherein the at least one search configuration parameter defines criteria for selecting samples to be used as the set of training candidates;
generate a first set of media content by labeling the set of training candidates based on the plurality of instances of the custom object label output by the at least one teacher model;
create a custom model using the at least one teacher model, wherein the custom model is a machine learning model trained using the first set of media content;
obtain a subset of a second set of media content, wherein the subset of the second set of media content is selected based on outputs of the custom model as applied to the second set of media content, wherein the outputs of the custom model include at least one first prediction for the subset of the second set of media content; and
apply an advanced machine learning model to the obtained subset of the second set of media content, wherein a domain45 used by the custom model is a subset of a domain used by the advanced machine learning model
, wherein the domain of each of the custom model and of the advanced machine learning model is a set of potential features used by the respective model, wherein the set of potential features of the domain used by the custom model is a subset of the set of potential features used by the advanced machine learning model.
Claim 14 is rejected similar to claim 2:
Re 14. Wei of the combination of Wei,de Silva,Bilenko teaches The system of claim 13, wherein the system is further configured to:
train the custom model using the at least one teacher model, wherein a domain used by the custom model is a subset of a domain used by the at least one teacher model; and
send, from a second system to a first system, the trained custom model for deployment as the first machine learning model at the first system, wherein the first system is remote from the second system.
Claim 15 is rejected similar to claim 3:
Re 15. Wei of the combination of Wei,de Silva,Bilenko teaches The system of claim 14, wherein the system is further configured to:
label at least a portion of the plurality of training candidates with the custom object label based on the output plurality of instances of the custom object label in order to create labeled media content; and
generate the set of second media content based on the labeled media content.
Claim 16 is rejected similar to claim 4:
Re 16. Wei of the combination of Wei,de Silva,Bilenko teaches The system of claim 15, wherein the custom model is configured to detect custom objects (“custom model” “data objects (such as video files 318)”,c.9,ll. 23-25 & c.7,ll. 25-30) corresponding to the custom object label when applied to features of media content.
Claim 19 is rejected similar to claim 7:
Re 19. Wei of the combination of Wei,de Silva,Bilenko teaches The system of claim 15, wherein the second set of media content includes exactly one sample of previously labeled media content.
Claim 20 is rejected similar to claim 8:
Re 20. Wei of the combination of Wei,de Silva,Bilenko teaches The system of claim 15, wherein the system is further configured to:
select the plurality of training candidates from among a third set of media content based on at least one search parameter, wherein the at least one search parameter defines criteria for identifying the third set of media content and for selecting samples from among the third set of media content to be used as the training candidates wherein the set of third media content includes a plurality of portions of media content showing a custom object corresponding to the custom object label.
Claim 23 is rejected similar to claim 11:
Re 23. Wei of the combination of Wei,de Silva,Bilenko teaches The system of claim 13, wherein each of the first machine learning model and a second machine learning model is a classifier, wherein the first machine learning model is configured to output a plurality of first classes, wherein the second machine learning model is configured to output a plurality of second classes, wherein the plurality of first classes is a subset of the plurality of second classes, wherein each of the plurality of first classes and the plurality of second classes includes a custom object class corresponding to the custom object label.
Claim(s) 5 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES):
PNG
media_image15.png
930
558
media_image15.png
Greyscale
Re 5., the combination of Wei and de Silva teaches not The method of claim 4, wherein the custom model (Wei: fig. 1:142 via a “customer”, c.8,l. 66 to c.9,l. 6 as modified by the combination) is a text-to-visual content model trained to detect custom objects corresponding to the custom object label (“for use in creating a custom model”, Wei: c.9,ll. 25-30, via fig. 3:309: “USER”) and to detect at least one predetermined (video) object for which the at least one teacher model (142,146 as modified via the combination) is trained to detect (local & global defects via 35 USC 112(f), not invoked:
The combination of Wei and de Silva does not teach the difference of claim 5:
--a text-to-visual content model trained to detect custom objects … and to detect at least one predetermined object for which the at least one teacher model is trained to detect--.
Acharya teaches the difference of claim 5:
5. The method of claim 4, wherein the custom model (“supervised by a large readily available pre-trained teacher model.”, pg. 13, 3rd para, penult S) is a text-to-visual content (“computer vision”, pg. 24, 2.2.3.1 Computer Vision 2nd S) model trained (via “knowledge distillation” pg. 24, 2.2.3.1 Computer Vision 1st S:pg. 24: fig. 2.1: “Generic KD Framework”) to detect custom objects (via “object detection” pg. 24, 2.2.3.1 Computer Vision 3rd S, 3rd listed item) corresponding to the custom object (“regularized”, pg. 13, 1st para, 1st S) label and to detect at least one predetermined object (via said objection detection) for which the at least one teacher model (fig. 1.4: “Teacher Model”: “Pre-Trained”) is trained to detect (an object).
MPEP 2143 Examples of Basic Requirements of a Prima Facie Case of Obviousness [R-01.2024]
I. EXAMPLES OF RATIONALES
C. Use of Known Technique To Improve Similar Devices (Methods, or Products) in the Same Way
To reject a claim (5) based on this rationale, Office personnel must resolve the Graham factual inquiries. Then, Office personnel must articulate the following:
a finding that the prior art (Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) as applied to claims (1,2,3,4,7,8,11 and 12 and 13,14,15,16,19, 20, 23) above further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES)) contained a "base" device (method, or product-machine learning:“(e.g., deep) machine learning”, Wei, c.2,ll. 19-22;“deep learning”46, Acharya, pg. 12, 1.1 Background, 3rd para, 2nd S) upon which the claimed invention47 can be seen as an "improvement;"
a finding that the prior art (Archarya) contained a "comparable" (teaching-student) device (method, or product that is not the same as the base device) that has been improved in the same way (“maintaining a high accuracy of detection performance”, Wei, c.2,ll. 50-55; “the large deep model’s logits output…to precisely replicate the teacher model’s final prediction”, pg. 14, 1.2.1 Knowledge & pg.14: 1.2.1.1 Response-Based Knowledge, 2nd S ) as the claimed invention48;
(3) a finding (Archarya, pg. 19, 1.5 Significance Study, 2nd S) that one of ordinary skill in the art could have applied the known (teacher-student) "improvement" technique in the same way to the "base" (machine-learning) device (method, or product) and the results would have been predictable to one of ordinary skill in the art (such as expecting accurate deep learning teacher final predictions):
PNG
media_image16.png
1687
876
media_image16.png
Greyscale
; and
(4) whatever additional findings (no additional findings) based on the Graham factual inquiries may be necessary, in view of the facts of the case under consideration, to explain a conclusion of obviousness.
The rationale to support a conclusion that the claim would have been obvious is that a method of enhancing a particular class of devices (methods, or products) has been made part of the ordinary capabilities of one skilled in the art based upon the teaching of such improvement in other situations. One of ordinary skill in the art would have been capable of applying this known method of enhancement to a "base" device (method, or product) in the prior art and the results would have been predictable to one of ordinary skill in the art. "It's enough … to show that there was a known problem … in the art, that [another reference] … helped address that issue, and that combining the teachings of [the two references] wasn't beyond the skill of an ordinary artisan. Nothing more is required to show a motivation to combine under KSR." See Intel Corp. v. PACT XPP Schweiz AG, 61 F.4th 1373, 1380-81, 2023 USPQ2d 297 (Fed. Cir. 2023) (finding that both prior art references "address the same problem and that [the secondary reference’s] cache was a known way to address that problem is precisely the reason that there's a motivation to combine under KSR and our precedent.").
The Supreme Court in KSR noted that if the actual application of the technique would have been beyond the skill of one of ordinary skill in the art, then using the technique would not have been obvious. KSR, 550 U.S. at 417, 82 USPQ2d at 1396. If any of these findings cannot be made, then this rationale cannot be used to support a conclusion that the claim would have been obvious to one of ordinary skill in the art.
Claim 17 rejected similar to claim 5:
17. The system of claim 16, wherein the custom model is a text-to-visual content model trained to detect custom objects corresponding to the custom object label and to detect at least one predetermined object for which the at least one teacher model is trained to detect.
Claim(s) 6 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of ACHARYA (EVALUATE THE EFFECTIVENESS OF VARIOUS KNOWLEDGE DISTILLATION METHODS FOR SMALLER NEURAL NETWORK ARCHITECTURES) as applied in claims 5 and 17 further in view of YIN (CN 108133476 B) with Google Patents machine translation:
PNG
media_image17.png
930
640
media_image17.png
Greyscale
Re 6, the combination of Wei and de Silva and Acharya teaches The method of claim 5, wherein the at least a portion of the labeled media content (Wei: fig. 3:318,320, 322: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”; “LABELED DEFECTS”) is below a threshold proportion of a total number of the plurality of training candidates (Wei: in fig. 3:318,320: “VIDEO FILE(S) (E.G., AND OTHER DATA)”: “TRAINING VIDEO FILE(S)”).
The combination of Wei and de Silva and Acharya does not teach the difference of claim 6: “a threshold proportion”.
YIN teaches the difference of claim 6:
6. The method of claim 5, wherein the at least a (“pixel”49, pg. 11, 1st txt blk) portion of the labeled (or “marked”50, pg. 8, 4th txt blk: last S, video display as a label) media content (i.e., “videos”, pg. 9, 6th txt blk-pixels) is (“otherwise”, pg. 13, 4th text blk) below a (“set”, pg. 13, 4th txt blk) threshold (“value”, pg. 13 4th txt blk) proportion (or “probability”51, pg. 13, 4th blk) of a total (“pixel”, pg. 9, 4th txt blk) number of the plurality of training candidates52 (or trained candidate objects via “trained” “candidate lung nodule image information”53, pg. 9, 4th txt blk).
Since Wei suggests the option of labeling training data, c.9,ll.1-5, one of skill in the art of labeling can make the combination of Wei’s be as Yin’s predictably recognizing the change providing an expected or looked forward to “detection scheme with higher accuracy”.
Claim 18 is rejected similar to claim 6:
18. The system of claim 17, wherein the at least a portion of the labeled media content is below a threshold proportion of a total number of the plurality of training candidates.
Claim(s) 9 and 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of Gasper et al. (US 2022/0327119 A1):
PNG
media_image18.png
930
640
media_image18.png
Greyscale
Re 9., the combination of Wei and de Silva teaches not The method of claim 1, further comprising:
enriching the subset (i.e., any one subset) of first media content based on the plurality of second (defect-detector) predictions (146) to create a set of enriched media content; and
sending (via fig. 1: arrows) the set of enriched media content to be used for populating a dashboard (“dashed borders”, c.23,ll. 25-30).
The combination of Wei and de Silva does not teach the difference of claim 9 of:
--enriching the subset …based on the plurality of second predictions to create a set of enriched media content; and
sending the set of enriched media content to be used for populating a dashboard--.
Gasper teaches the difference of claim 9:
9. The method of claim 1, further comprising:
enriching the (“attribute” [0051]) subset of first media content based on the plurality of second (“match group” [0052] 3rd S) predictions to create a set (or “generate…datasets and portions thereof” [0052] 1st S & last S: fig. 1:106: “Content Graph Portion”) of enriched media content (“as enrichment data 147b” [0053] 4th S: figs. 1,13:147b,1347b: “Enrichment Data”); and
sending the set (106) of enriched (“YouTube” “video” [0043]) media content to be used for populating a dashboard (fig. 13:1313: “Dashboard/Visualization Data”).
MPEP 2143 Examples of Basic Requirements of a Prima Facie Case of Obviousness [R-01.2024]
I. EXAMPLES OF RATIONALES
F. Known Work in One Field of Endeavor May Prompt Variations of It for Use in Either the Same Field or a Different One Based on Design Incentives or Other Market Forces if the Variations Are Predictable to One of Ordinary Skill in the Art
To reject a claim (9) based on this rationale, Office personnel must resolve the Graham factual inquiries. Then, Office personnel must articulate the following:
(1) a finding that the scope and content of the prior art [Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) as applied to claims (1,2,3,4,7,8,11 and 12 and 13,14,15,16,19, 20, 23) above further in view of Gasper et al. (US 2022/0327119 A1)], whether in the same field (same content processing):
applicant’s disclosure [002]:
--TECHNICAL FIELD
[002] The present disclosure relates generally to processing content such as images using machine learning, and more specifically to training and using machine learning models for tracking custom objects.—
Wei (US 11,445,168):
--BACKGROUND
Generally described, computing devices utilize a communication network, or a series of communication networks, to exchange data. Companies and organizations operate computer networks that interconnect a number of computing devices to support operations or provide services to third parties. The computing systems can be located in a single geographic location or located in multiple, distinct geographic locations (e.g., interconnected via private or public communication networks). Specifically, data centers or data processing centers, herein generally referred to as “data centers,” may include a number of interconnected computing systems to provide computing resources to users of the data center. The data centers may be private data centers operated on behalf of an organization or public data centers operated on behalf, or for the benefit of, the general public. Service providers or content creators (such as businesses, artists, media distribution services, etc.) can employ one or more data centers to deliver content (such as web sites, web content, or other digital data) to users or clients.
Gasper:
FIELD
[0002] Various embodiments relate generally to data science and data analysis, computer software and systems, and data-driven control systems and algorithms based on graph-based data arrangements, among other things, and, more specifically, to a computing platform configured to receive and analyze datasets to implement a data model with which to identify relevant data catalog data derived from graph-based data arrangements, whereby a processor may be configured to cause implementation of subsets of programmatic executable instructions based on relevant data catalog54 data to perform an action (e.g., autonomously, semi-autonomously, or manually), at least one example of which may be configured to generate a concept interface portion relevant to a concept of interest, among other things.
of endeavor as that of the applicant’s invention or a different field of endeavor, included a similar or analogous (machine learning) device (method, or product);
(2) a finding that there were design incentives or market forces which would have prompted adaptation of the known (machine learning) device (method, or product):
Wei (US 11,445,168), c.1, ll.5-25: a business:
--BACKGROUND
Generally described, computing devices utilize a communication network, or a series of communication networks, to exchange data. Companies and organizations operate computer networks that interconnect a number of computing devices to support operations or provide services to third parties. The computing systems can be located in a single geographic location or located in multiple, distinct geographic locations (e.g., interconnected via private or public communication networks). Specifically, data centers or data processing centers, herein generally referred to as “data centers,” may include a number of interconnected computing systems to provide computing resources to users of the data center. The data centers may be private data centers operated on behalf of an organization or public data centers operated on behalf, or for the benefit of, the general public. Service providers or content creators (such as businesses, artists, media distribution services, etc.) can employ one or more data centers to deliver content (such as web sites, web content, or other digital data) to users or clients.
Gasper (US 2022/0327119 A1): marketing:
--[0004] While conventional approaches are functional, various approaches are not well-suited to significantly overcome the difficulties of data silos. Organizations, including enterprises, continue strive to understand, manage, and productively use large amounts of enterprise data. For example, consumers of enterprise organizations have different levels of skill and experience in using analytic data tools. Data scientists typically create complex data models using sophisticated analysis application tools, whereas other individuals, such as executives, marketing personnel, product managers, etc., have varying levels of skill, roles, and responsibilities in an organization. The disparities in various analytic data tools, reporting tools, visualization tools, etc., continue to frustrate efforts to improve interoperability and usage of large amounts of data.--;
(3) a finding that the differences between the claimed invention and the prior art were encompassed in known variations or in a principle known in the prior art:
Gasper (US 2022/0327119 A1) teaches the difference of claim 9:
9. The method of claim 1, further comprising:
enriching the (“attribute” [0051]) subset of first media content based on the plurality of second (“match group” [0052] 3rd S) predictions to create a set (or “generate …datasets and portions thereof” [0052] 1st S & last S: fig. 1:106: “Content Graph Portion”) of enriched media content (“as enrichment data 147b” [0053] 4th S: figs. 1,13:147b,1347b: “Enrichment Data”); and
sending the set (106) of enriched (“YouTube” “video” [0043]) media content to be used for populating a dashboard (fig. 13:1313: “Dashboard/Visualization Data”).;
(4) a finding that one of ordinary skill in the art, in view of the identified design incentives (said a business and marketing) or other market forces, could have implemented the claimed variation (of claim 9) of the prior art (Gasper (US 2022/0327119 A1)) by making Wei’s fig. 6:610: video-on-demand “VIRTUALIZATION SERVICES” be as Gasper’s fig. 13:1305: video-“content”, Gasper: [0045] 1st S, “Cloud Service(s)”:
PNG
media_image19.png
1269
1748
media_image19.png
Greyscale
, and the claimed variation (of claim 9) would have been predictable to one of ordinary skill in the art (of machine learning by expecting improved “accuracy”, Gasper [0203] last S, of classification of dashboard data: Gaspar: figs. 1:134a,13:1330: “Classifier”, “Dataset Ingestion Controller”); and
(5) whatever additional findings (no other additional finding are presented) based on the Graham factual inquiries may be necessary, in view of the facts of the case under consideration, to explain a conclusion of obviousness.
The rationale to support a conclusion that the claimed invention would have been obvious is that design incentives or other market forces could have prompted one of ordinary skill in the art to vary the prior art in a predictable manner to result in the claimed invention. If any of these findings cannot be made, then this rationale cannot be used to support a conclusion that the claim would have been obvious to one of ordinary skill in the art.
Claim 21 is rejected similar to claim 9:
21. The system of claim 13, wherein the system is further configured to:
enrich the subset of first media content based on the plurality of second predictions to create a set of enriched media content; and
send the set of enriched media content to be used for populating a dashboard.
Claim(s) 10 and 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wei et al. (US 11,445,168 B1) in view of da Silva et al. (US 2024/0111868 A1) further in view of Bilenko et al. (US 2013/0204809 A1), as applied in claims 24,25,26, as applied in claims 1,2,3,4,7,8,11 and 12 and 13,14,15,16,19,20,23 further in view of Rogers et al. (US 2021/0117859 A1):
PNG
media_image20.png
930
640
media_image20.png
Greyscale
Re 10., the combination of Wei and de Silva teaches not The method of claim 1, wherein the first machine learning model (142) is applied by an edge device deployed locally with respect to a source (“video”,c.3,ll.15-20) of the (provider’s 134) media content, wherein the second machine learning model (146) is deployed remotely from the edge device.
The combination of Wei and de Silva does not teach the difference of claim 10:
--edge device deployed locally with respect to…deployed remotely from the edge device--.
Rogers teaches the difference of claim 10:
10. The method of claim 1, wherein the first machine learning model (fig. 1:144: “Models”) is applied by an edge device (fig. 1:120: “Edge Server”) deployed (via “deploying a model” [0037] 1st S) locally (“obtained…and stored in a local directory” [0037] 3rd S) with respect to a source of the media content (fig. 1:160: “Media Source”), wherein the second machine learning model (fig. 1:128,144: “DL Model”, “Models”) is deployed remotely (via “a remote model store” [0038] 5th S: fig. 1:142: “Model Store”) from the edge device (120).
Since Wei suggests other “deployment ser-vices”, c.7,ll. 5-25, one of skill in the art of deployments can make the combination of Wei’s be as Rogers’ predictably recognizing the change reducing “latency”, Rogers [0020] last S.
Claim 22 is rejected similar to claim 10:
22. The system of claim 13, wherein the first machine learning model is applied by an edge device deployed locally with respect to a source of the media content, wherein the second machine learning model is deployed remotely from the edge device.
Conclusion
The prior art “nearest to the subject matter defined in the claims” (MPEP 707.05) made of record and not relied upon is considered pertinent to applicant's disclosure.
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
YU et al. (TW 1612488 B) with SEARCH machine translation
YU teaches a domain is (“i.e.,”) a potential feature set (fig. 4A:40: “KxN” block) in the context of a potential feature, pg. 11, 4th txt blk:
“the feature set in the L potential feature matrix 40, X .sub.S is the feature set reconstructed after encoding and decoding, x .sub.T is the feature set of the target domain (ie, the feature set of the new product) ”
as the closest to the claimed “the domain…is a set of potential features” of claim 24.
DANIELLA et al. (KR 20210078488 A) with SERACH machine translation
DANIELLA teaches, pg. 13, 3rd txt blk:
“That is, at some levels there may be multiple domains that may individually qualify as a feature of interest, and based on the plurality of traits, the highest likelihood or probability that any of the potential features will be characterized It is only possible to determine which is an actual feature of interest only if it is determined that it has.”
as the closest to the claimed “the domain…is a set of potential features” of claim 24.
Gilboa et al. (US 11,954,926 B2) corresponding to
DANIELLA et al. (KR 20210078488 A) with SERACH machine translation
Gilboa teaches:
“That is, there could be multiple areas55 which could individually qualify as a feature of interest on some levels and only in determining which of the potential features has the highest likelihood or probability of being the feature on the basis of a plurality of characteristics, is it possible to determine which is the actual feature of interest.”
as the closest to the claimed “the domain…is a set of potential features” of claim 24.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DENNIS ROSARIO whose telephone number is (571)272-7397. The examiner can normally be reached Monday-Friday, 9AM-5PM EST.
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.
/DENNIS ROSARIO/Examiner, Art Unit 2676
/Henok Shiferaw/Supervisory Patent Examiner, Art Unit 2676
1 output: Computers. to transfer (information) from internal storage to an external medium, wherein storage is defined: Computers. memory ( def 11 ), wherein memory is defined: Also called computer memory, Computers. the capacity of a computer to store information subject to recall (Dictionary.com)
2 machine learning Computers, Digital Technology. the capacity of a computer to process and evaluate data beyond programmed algorithms, through contextualized inference (often used attributively).
3 underlying detection function y(x)=y(configured), wherein x=configured is a single act and thus not acts
4 underlying detection function y(x)=y(trained), wherein x=trained is a single act and thus not acts
5 underlying detection function y(x)=y(trained), wherein x=trained is a single act and thus not acts
6 underlying detection function y(x)=y(trained), wherein x=trained is a single act and thus not acts
7 BROAD WORD: domain: a region characterized by a specific feature, type of growth or wildlife, etc.. (Dictionary.com): most consistent meaning in view of applicant’s disclosure of potential feature/value
8 area: any particular extent of space or surface, wherein space is defined: the unlimited or incalculably great three-dimensional realm or expanse in which all material objects are located and all events occur, wherein realm is defined: the region, sphere, or domain within which anything occurs, prevails, or dominates. (Dictionary.com)
9 subset: a set that is a part of a larger set. (Dictionary.com)
10 categorize: to describe by labeling or giving a name to; characterize. (Dictionary.com)
11 in: preposition (adverb): a modifier of verbs (none presented), nouns (“feature”), or adjectives (“potential”) (Dictionary.com)
12 i.e., that is; that is to say; in other words. (Dictionary.com)
13 in: another prepositional modifier further limiting “domain”
14 train: intr to do exercises and prepare for a specific purpose the athlete trained for the Olympics; wherein exercise is defined: noun a set of movements, questions, tasks, etc, designed to train, improve, or test one's ability in a particular field, wherein train is defined: tr to guide or teach (to do something), as by subjecting to various exercises or experiences (Dictionary.com)
15extract: to take or copy out (matter), as from a book, wherein take is defined: (of a person or thing) to win favor or acceptance, wherein favor is defined: preferential treatment (Dictionary.com)
16 BROAD CLAIM LANGUGAE: domain: A.a field of action, thought, influence, etc. B.a realm or range of personal knowledge…etc.C.a region characterized by a specific feature… etc. (Dictionary.com)
17 BROAD CLAIM LANGUAGE: Regarding “of”, wherein “of” is defined: (used to indicate possession, connection, or association) (Dictionary.com), via applicant’s disclosure:
[0099]AII examples and conditional language recited herein are intended for pedagogical purposes to aid the reader in understanding the principles of the disclosed embodiment and the concepts contributed by the inventor to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the disclosed embodiments, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure.
However, improperly reading limitations from applicant’s disclosure in claim 24:
“potential features”=”domain” as further discussed below: THE CLAIMED INVENTION AS A WHOLE
18 see above-“of” via applicant’s disclosure-footnote
19 see above-“of” via applicant’s disclosure-footnote
20 see above-“of” via applicant’s disclosure-footnote
21 SEARCH: “potential features” may be a term of art.
22 see above-“of” via applicant’s disclosure-footnote
23 see above-“of” via applicant’s disclosure-footnote
24 see above-“of” via applicant’s disclosure-footnote
25 THE CLAIMED INVENTION AS A WHOLE regarding “potential features” (“i.e.,” “the domain”: [0056]: “domain” has multiple re-wordings (i.e.,) through-out the disclosure):
The problem in applicant’s disclosure is:
[003] With the rapid adoption of computerized monitoring technologies, the amount of media content being captured and processed has exploded in recent years. With this explosion of media content that needs to be analyzed and processed, the need for efficient ways to process this vast amount of data is more acute than ever. Video monitoring technologies are being used for many different purposes, such as home monitoring (e.g., video doorbells which watch for activity outside of a door), vehicle monitoring (e.g., systems which monitor video for parking or other vehicle-related violations), hospitality (e.g., video monitoring inside of hotels), and many more. Although software can be installed locally where the media content is captured, processing such large amounts of media content presents a challenge and not all sites are equipped with the computing resources to handle such processing.
Wei does not appear to teach this problem: an indication of non-obviousness.
The solution is (briefly):
[006] Solutions which would more efficiently process [see fig. 3] large volumes of media content are therefore highly desirable. It would further be beneficial for such solutions to allow for custom definitions of objects in order to facilitate monitoring with respect to objects which are not already known to the relevant systems.
I see this custom aspect in claim 24 (so that is an indication of non-obviousness); however, I don’t see fig. 3:322: ”MA” (“which is configured to analyze the outputs of the advanced model 331 as metadata for the respective interesting portions of content identified by the custom model 321.” [0052]) in claim 24 (an indication of OBVIOUSNESS).
Since “domain” is re-worded multiples times in applicant’s disclosure, the role of “potential features” (i.e., domain) is not clear in the disclosed solution and appears to require an in-depth grammatical analysis of the disclosed “domain” (identified as broad claim language) and “potential features” to fully see/understand (the broadest reasonable interpretation) what is meant by “domain” and “potential features” (an indication of OBVIOUSNESS).
This absence and no clear (i.e., broad) language of “domain” (i.e., the claimed “potential features”) in claim 24 of the disclosed solution is an indication of OBVIOUSNESS.
35 USC 103 obviousness meter: I got 3 OBVIOUSNESS vs. 2 non-obviousness.
REMEMBER limitations (such as “domain”=”potential features”) from applicant’s disclosure are not read into claim 24.
26 (italics) represent claim limitations already taught
27 (italics) represent claim limitations already taught
28 label: a word or phrase indicating that what follows belongs in a particular category or classification (Dictionary.com)
29 label: a piece of paper, card, or other material attached to an object (or defect) to identify it or give instructions or details concerning its ownership, use, nature, destination, etc; tag (Dictionary.com)
30 extract (used with object): to take or copy out (matter), as from a book, wherein take is defined: to pick from a number; select: (Dictionary.com)
31 student: any person who studies, investigates, or examines thoughtfully, wherein person is defined: an individual of distinction or importance, wherein individual is defined: a distinct, indivisible entity; a single thing, being, instance, or item (Dictionary.com)
32parameter: Statistics. a variable entering into the mathematical form of any distribution such that the possible values of the variable correspond to different distributions, wherein value is defined: Mathematics. magnitude; quantity; number represented by a figure, symbol, or the like, wherein quantity is defined: an exact or specified amount or measure, wherein measure is defined: (7) any standard of comparison, estimation, or judgment (Dictionary.com)
33 i.e., subset set
34 THE CLAIMED INVENTION AS A WHOLE regarding the claimed “creating a custom model using the at least one teacher model”:
The large data problem is discussed in the rejection of claim 24.
The solution is (briefly & further detailed: like a student trained on subset (elementary school workbook) referring back to a teacher (college textbook) for help):
[006] Solutions which would more efficiently process [see fig. 3] large volumes of media content are therefore highly desirable. It would further be beneficial for such solutions to allow for custom definitions of objects in order to facilitate monitoring with respect to objects which are not already known to the relevant systems.
[0029]The disclosed embodiments may be utilized for image processing in situations where a large amount of image data is continuously collected in order to perform basic analysis of those images using the custom model deployed at an edge device and to only perform advanced analyses of the images when the basic analysis yields predictions indicating that those images are potentially interesting, for example, when the custom model outputs predictions indicating that certain images show the custom object. To this end, in accordance with various disclosed embodiments, the custom model may be trained on a subset of the feature domain used by the teacher models such that applying the custom model created as described herein requires less processing than applying teacher models or other more advanced models. Accordingly, the disclosed embodiments can reduce total processing of content by limiting the amount of analysis performed using heavier models (e.g., models which have larger feature domains, more granular analysis, more kinds of outputs, otherwise require more processing, etc.) as enabled by a lighter model.
I see custom: indication of non-obviousness; however, I don’t see in claim 1: the custom model may be trained on a subset of the feature domain. This absence indicates obviousness in claim 1.
35 italics represent claim limitations already taught
36 label: to mark with a label, wherein mark is defined: tr to select, designate, or doom by or as if by a mark (Dictonary.com)
37 video: a program, movie, or other visual media product featuring moving images, with or without audio, that is recorded and saved digitally or on videocassette (Dictionary.com)
38 italics represent claim limitations already taught
39 “subset” is part of applicant’s disclosed large data solution, but this “subset” is not clearly in the context of training.
40 “subset” is part of applicant’s disclosed large data solution, but this “subset” is not clearly in the context of training.
41 “subset” is part of applicant’s disclosed large data solution, but this “subset” is not clearly in the context of training.
42 “subset” is part of applicant’s disclosed large data solution, but this “subset” is not clearly in the context of training.
43 “subset” is part of applicant’s disclosed large data solution, but this “subset” is not clearly in the context of training.
44 “subset” is part of applicant’s disclosed large data solution, but this “subset” is not clearly in the context of training.
45 THE CLAIMED INVENTION AS A WHOLE: These subsets/domain is not clearly in the context of training
46 deep learning: an advanced type of machine learning that uses multilayered neural networks to establish nested hierarchical models for data processing and analysis, as in image recognition or natural language processing, with the goal of self-directed information processing. (Dictionary.com)
47 Applicant’s disclosure:
[0027]The disclosed embodiments allow for custom defining objects which may not be represented in the initial training data used to initially configure the teacher machine learning model. This, in turn, allows users to define custom classes to be utilized by machine learning models without requiring explicit programming or otherwise providing a large amount of text for training. Moreover, various disclosed embodiments provide techniques which obtain content to be used for training which does not require multiple initial samples showing the custom-defined objects in order to train the resulting machine learning models to accurately identify the custom-defined objects.
48 Applicant’s disclosure:
[0027]The disclosed embodiments allow for custom defining objects which may not be represented in the initial training data used to initially configure the teacher machine learning model. This, in turn, allows users to define custom classes to be utilized by machine learning models without requiring explicit programming or otherwise providing a large amount of text for training. Moreover, various disclosed embodiments provide techniques which obtain content to be used for training which does not require multiple initial samples showing the custom-defined objects in order to train the resulting machine learning models to accurately identify the custom-defined objects.
49 pixel: Computers, Television. the smallest element of an image that can be individually processed in a video display system. (Dictionary.com)
50 mark: to furnish with figures, signs, tags, etc., to indicate price, quality, brand name, or the like, wherein tags is defined: a descriptive word or phrase applied to a person, group, organization, etc., as a label or means of identification (Dictionary.com)
51probability: statistics a measure or estimate of the degree of confidence one may have in the occurrence of an event, measured on a scale from zero (impossibility) to one (certainty). It may be defined as the proportion of favourable outcomes to the total number of possibilities if these are indifferent (mathematical probability ), or the proportion observed in a sample ( empirical probability ), or the limit of this as the sample size tends to infinity ( relative frequency ), or by more subjective criteria ( subjective probability ) (Dictionary.com)
52 candidate: a person or thing regarded as suitable or likely for a particular fate or position, wherein thing is defined: an object, fact, affair, circumstance, or concept considered as being a separate entity (Dictionary.com)
53 information Computers. important or useful facts obtained as output from a computer by means of processing input data with a program, wherein facts is defined: something known to exist or to have happened, wherein thing is defined: a material object without life or consciousness; an inanimate object (Dictionary.com)
54 catalog: a list of the contents of a library or a group of libraries, arranged according to any of various systems. (Dictionary.com)
55 Wei (US 11,445,168 B1) as applied in the rejection of claim 24 also teaches area (video-frame).