DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 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 1–20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more, rendering the claims ineligible for patent protection.
I. Alice Step 1:
Claims 1, 11, and 16 are directed to the abstract idea of mathematical modeling and organizing/comparing information. Mathematical Concepts: The claims recite the use of "theoretical manifolds" that are "multidimensional" and "parameterized by a plurality of features". This describes a mathematical framework for representing model performance. Identifying portions of a manifold based on "theoretical geometry” to denote an "expected manner of performance" is a purely mathematical exercise. Mental Processes/Organizing Information: The core of the claim involves: Receiving an aspect of a region. Linking that aspect to features. Comparing performance metrics. Making a recommendation. These steps constitute "collecting information, analyzing it, and presenting the results," which are long-standing fundamental mental processes. The addition of "visual aspect" and "physical region" in the amendments does not change the fundamental nature of the process, as the "linking" and "identifying" still occur within the abstract mathematical space of the performance manifold.
II. Alice Step 2:
The claims do not recite "significantly more" to transform the abstract idea into a patent-eligible application. Generic Computer Components: The instructions are performed by a "hardware processor" and a "computing system". The specification confirms these are "processing resources" and "generic computer system components" used to implement the approach. Performing an abstract idea on a generic computer is insufficient for eligibility.
Lack of Technical Improvement: While the specification argues that the invention provides a "technical solution" for comparing model performance, the claims do not recite a specific improvement to computer functionality itself. Instead, they describe using a computer as a tool to perform a new mathematical analysis (the manifold comparison) to arrive at a better recommendation. Pre-existing Techniques: Claim 7 and 15 recite "running a residual network feature extractor". The specification acknowledges that these extractors (e.g., ResNet50) are known algorithms. Using a well-known AI tool to feed data into an abstract mathematical model does not constitute an "inventive concept." Insignificant Pre-solution Activity: Receiving "images" or "attributes” to define the region of interest is merely data gathering (pre-solution activity) which does not impart eligibility.
The amendments adding "visual aspect," "physical region," and "image features" merely limit the abstract idea to a particular technological field (geospatial/image modeling). Under settled law, field-of-use limitations or the recitation of "physical" inputs do not save a claim that is fundamentally directed to an abstract mathematical and mental process. Consequently, claims 1–20 are rejected under 35 U.S.C. 101.
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.
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) 1-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over VINCENT (US20170061242A1) in view of BRAND (US20040091152A1).
Regarding claim 1, VINCENT teaches a method of recommending at least one prediction model from among a plurality of different prediction models, each one of the different prediction models having been trained based on a respective training data set, each one of the different prediction models being operable on a set of inputs that at least partially defines a respective theoretical manifold, each of the theoretical manifolds being multidimensional and being parameterized by a plurality of different image features, the method comprising:
receiving an aspect of a physical region definable in relation to the theoretical manifolds of the different prediction models (VINCENT discloses generating matrices of eigenvectors and covariance matrices to define modes of variation (a multidimensional space/manifold) that are parameterized by shape parameters and grey-level appearance parameters (image features)
for each of the different prediction models (VINCENT, fig. 3, para. 0096-117),
linking the aspect of the physical region to at least some of the image features parameterizing the respective manifold (VINCENT discloses initialising the model fitting in the image based on corresponding points, using a point correspondence that indicates a relationship between the points in the image and points in the models); and
identifying one or more portions of the respective manifold based on the image features determined by the linking, the one or more portions having a theoretical geometry that denotes an expected manner of performance of the respective model for the visual aspect of the physical region (VINCENT teaches evaluating a model's fit using a "predetermined criterion" which can comprise a "geometric criterion," such as a maximum distance or rotation (theoretical geometry), to classify whether the fit is a "good fit" or "bad fit" (expected manner of performance));
comparing the expected manners of performance of the different prediction models for the visual aspect of the physical region (VINCENT teaches comparing each model's individual fit against the geometric criterion to decide if it is acceptable, and combining the acceptable models using weights based on spatial distance); and
based on the comparison, recommending one or more of the different prediction models (VINCENT, fig. 5, S17, acceptable models, para. 0113).
VINCENT is silent to teaching a visual aspect of a physical region.
In the same field of endeavor, BRAND teaches a visual aspect of a physical region (BRAND teaches mapping a high-dimensional measured sample (such as a digital image of a human face) to low-dimensional coordinates on the manifold, para. 0030-40).
Therefore, it would have been obvious to an ordinary artisan seeking to optimize multi-model computer vision tasks would be strongly motivated to combine the teachings of Vincent and Brand. Vincent provides an outstanding methodology for breakdown and feature matching using localized sub-model sets to improve tracking precision. However, Vincent notes a persistent challenge: if a statistical model is trained on a limited dataset, it can easily fail to capture a true indication of acceptable variation when processing unseen images. Brand directly solves this specific data-science vulnerability by offering an elegant, closed-form mathematical framework. Brand demonstrates that treating high-dimensional image variations as an embedded, low-dimensional theoretical manifold dramatically reduces information loss, filters out sampling noise perpendicular to the data surface, and mathematically formalizes relationships between localized charts. Integrating Brand’s manifold mapping into Vincent’s hierarchical model architecture would provide the exact mathematical framework required to objectively evaluate, smooth, and recommend the best performing model configurations for any received input region. The combination relies entirely on standard linear algebra and statistical tools applied to computer vision, yielding completely predictable results.
Claims 2, 3, 12, 13, 17, 18 (Manifold Representation and Sub-Models)
Regarding Claims 2, 3, 12, 13, 17, 18 are rejected under 35 U.S.C. 103 as unpatentable over VINCENT in view of BRAND. These claims require generating a representation of the manifold by determining strongly correlated features, creating a plurality of sub-models that together approximate the manifold, and defining the representation using those sub-models. Vincent explicitly discloses an "arbitrary decomposition of the object" into a plurality of models that "together model a region of interest," where localized models overlap to represent the global space. Vincent selects these subsets based on feature correlation, such as targeting regions of high curvature. Brand explicitly translates this concept into manifold geometry by decomposing the high-dimensional data space into "locally linear low-dimensional patches or 'charts'" that are subsequently merged into a single coordinate system ("connection") to approximate the global manifold. Brand fits these localized sub-models using a Gaussian Mixture Model (GMM) where the dominant axes span the strongly correlated linear features of the manifold. Combining Vincent's localized image model patching with Brand's mathematical chart-tessellation mapping to approximate an intractable manifold is structurally identical to the claimed steps.
Claims 4, 5, 14, 19, 20 (Prototype Exemplars)
Regarding Claims 4, 5, 14, 19, 20 are rejected under 35 U.S.C. 103 as unpatentable over VINCENT in view of BRAND. These claims require generating prototype exemplars (objects, images, or image collections) for each created sub-model, which characterize the geometry of a portion of the manifold. Vincent teaches using a "large number of examples of a good fit... and a bad fit... in a training stage" to allow the system to learn characteristics of specific model areas based on residual error. Brand explicitly teaches that for each localized sub-model/chart, the system handles data points acting as explicit structural examples. Crucially, Brand teaches a reverse/back-projection mapping that allows the system to "synthesize" novel, distinct sample images from any coordinate point on a chart. These synthesized images serve precisely as the claimed "prototype exemplars" to characterize the surface, angles, and geometry of that local slice of the manifold. It would be obvious to generate these synthetic examples using Brand's reverse-mapping equations to evaluate and visualize individual chart constraints.
Claims 6 & 7, 15 (Determining Strongly Correlated Features)
Regarding Claims 6, 7, and 15 are rejected under 35 U.S.C. 103 as unpatentable over VINCENT in view of BRAND. Claim 6 specifies determining strongly correlated features via a user-specified list. Vincent teaches that landmarks or points of interest can be "manually marking corresponding predetermined points" by an expert based on domain knowledge. Selecting features via a manual, user-specified target list is a notoriously old and well-known industry standard for initialization. Claims 7 and 15 require running a residual network (ResNet) feature extractor. While Vincent and Brand predate the ubiquity of modern deep-learning ResNet architectures, both explicitly teach automating feature extraction. Vincent uses automated grey-level and shape parameter tracking, while Brand teaches automated "kernel-based dimensionality reduction" and feature selection. At the time of the invention, replacing traditional hand-crafted kernel extractors with a standard, off-the-shelf convolutional neural network (specifically, a generic Residual Network like ResNet50) to automate image feature extraction was a matter of routine engineering configuration yielding predictable results.
Claims 8, 9, 10 (Input and Region Specifications)
Regarding Claims 8, 9, and 10 are rejected under 35 U.S.C. 103 as unpatentable over VINCENT in view of BRAND. These claims limit the data to geospatial/geotemporal data (Claim 8), parameterizing attributes (Claim 9), or a plurality of images (Claim 10). Vincent teaches processing sets of anatomical volume slices or annotated multi-dimensional surface images (a plurality of images), where regions are defined mathematically as shape or appearance attributes. Brand notes that its manifold-mapping technique is entirely general-purpose, applicable to any "high-dimensional bio-mechanical data," digital images, or "physical system" measurements. Applying these generic mathematical steps to geospatial or geotemporal images is a mere field-of-use application of the prior art techniques, which is thoroughly obvious since geospatial satellite tracking relies on identical multi-spectral pixel intensity spaces.
Claims 11 & 16 (Computer-Readable Medium and System)
Regarding Claims 11 and 16 are rejected under 35 U.S.C. 103 as unpatentable over VINCENT in view of BRAND. These claims recite a non-transitory computer-readable medium and a hardware system comprising at least one processor and memory to execute the steps of Claim 1. Vincent explicitly discloses a hardware "computer 1" comprising a "CPU 1a," "volatile memory 1b (RAM)," and "non-volatile storage in the form of a hard disk drive 1c" configured to read tangible instructions and execute the image processing steps.
Response to Arguments
Applicant’s arguments with respect to claim(s) 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WEN WU HUANG whose telephone number is (571)272-7852. The examiner can normally be reached Mon-Fri 10-6.
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, Wesley Kim can be reached at (571) 272-7867. 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.
/WEN W HUANG/ Primary Examiner, Art Unit 2648
Prosecution Insights