DETAILED ACTION
Response to Amendment
Claims 1-13 were previously pending. Applicant’s amendment filed November 21, 2025, has been entered in full. Claims 1, 6, 7, 10, and 11 are amended. Claims 4, 5, 9, and 12 are cancelled. No new claims are added. Accordingly, claims 1-3, 6-8, 10-11, and 13 are now pending.
Response to Arguments
Applicant traverses the previous written description rejections under 35 U.S.C. 112(a) (Remarks filed November 21, 2025, hereinafter Remarks: Pages 6-9). Examiner respectfully disagrees, as explained further below.
Applicant asserts that “the invention is not claiming an improvement in CNN architecture itself but in how object contours are mathematically represented, estimated, and reconstructed using elliptic Fourier descriptors.” (Remarks: Page 7). The declaration by Dávid Kiss filed on November 21, 2025, (hereinafter referred to as the Declaration) makes a similar statement (Declaration: statement 12). Examiner respectfully disagrees. Claim 1 is to a method with three steps, one of which is inputting an image to a trained neural network and another of which is performing estimating by the trained neural network. Claim 11 similarly is to a data processing system comprising a trained neural network. The neural network is plainly part of the claimed invention and must be adequately described in order to comply with the written description requirement under 35 U.S.C. 112(a).
Applicant points to seven disclosures in the specification (Remarks: Pages 7-8) and argues that “These disclosures together demonstrate clear possession of (1) the form of the representation, (2) the structure of the output vector, (3) the training targets, (4) the nature of the decoding/reconstruction step, and (5) the separate estimation of reference-contour descriptors and geometric transformation parameters” (Remarks: Page 8). These disclosures pertain to the outputs of the neural network, rather than how the neural network obtains those outputs. Even assuming arguendo that Applicant has shown possession of the outputs, that does not mean that they have described how the neural network obtains those outputs, which was the basis of the written description rejection.
Applicant next points to Mask R-CNN and FourierNet as prior art disclosures and argues that they render implementation details such as neural network structure, loss formulation, and training process so “routine and widely known” that they do not have to be described in order to show possession (Remarks: Pages 8-9). The Declaration includes similar statements (e.g., statements 10 and 11). Examiner respectfully disagrees.
Applicant’s assertions amount to an argument that one skilled in the art could write a program to achieve the claimed function. I.e., that one skilled in the art could follow and extrapolate the examples of Mask R-CNN and FourierNet to craft a neural network capable estimating the elliptic Fourier descriptor required by the claimed invention. However, “It is not enough that one skilled in the art could write a program to achieve the claimed function because the specification must explain how the inventor intends to achieve the claimed function to satisfy the written description requirement.” MPEP 2161.01, Subsection I. Mere knowledge of Mask R-CNN and FourierNet would not explain to one of ordinary skill in the art how the inventor intended to achieve the claimed use of a trained neural network to estimate an elliptic Fourier descriptor.
Indeed, neither of the Mask R-CNN or FourierNet examples cited by Applicant describes estimating an elliptic Fourier descriptor. As acknowledged in the specification (par. [0008], as-published), Mask R-CNN outputs a binary mask, rather than an elliptic Fourier descriptor. It is unclear how knowledge of a neural network that outputs a binary mask would explain to one of ordinary skill in the art how the inventor intended to use a trained neural network to estimate an elliptic Fourier descriptor, or would make the details of training and inference of such a network “routine and widely known”.
As acknowledged in the specification (par. [0013], as-published), FourierNet is described in the ‘Benbarka’ (“FourierNet: Compact mask representation for instance segmentation using differentiable shape decoders,” 2020) reference. Unlike the claimed invention, FourierNet does not estimate an elliptic Fourier descriptor. Instead, FourierNet takes a center of an object and measures distances from the center to the contour of the object at regular angular intervals (Benbarka: Sec. 3.1). The 1D sequence of distance measurements is represented by a Fourier transform (Benbarka: Sec. 3.1). It is unclear how knowledge of a neural network that outputs non-elliptic Fourier descriptors would explain to one of ordinary skill in the art how the inventor intended to use a trained neural network to estimate an elliptic Fourier descriptor, or would make the details of training and inference of such a network “routine and widely known”.
Furthermore, Examiner notes that the previously-cited ‘Xu’ prior art describes a Fourier descriptor that is very similar to the one in FourierNet. Like FourierNet, Xu describes taking a center of an object (Fig. 3 and Sec. 3.2.1, Inner center) and measuring distances from the center to the contour of the object at regular angular intervals (Sec. 3.2.1, Dense Contour Sampling). The 1D sequence of distance measurements may be represented by a Fourier transform (e.g., Sec. 3.2.3, Comparison with Other Fitting Methods). In their arguments traversing the previous rejection of claim 5 under 35 U.S.C. 103, Applicant acknowledges this 1D signature in Xu (Remarks: Page 16, last par.). Applicant also acknowledges that the previously-cited ‘Staib’ (“Boundary Fitting with Parametrically Deformable Models,” 1992) prior art “discloses elliptic Fourier descriptors” (Remarks: Page 16, last par.). Yet, Applicant then goes on to argue that it would not have been obvious for one of ordinary skill in the art to modify Xu to use the elliptic Fourier descriptor of Staib (Remarks: Page 16, last par.). In particular, Applicant argues:
“Staib does not teach or suggest replacing Xu’s one-dimensional radial signature f(θ) with a two-dimensional elliptic Fourier descriptor representation of x(t), y(t). Notably, combining Xu with Staib with Xu would require abandoning Xu’s fundamental use of a single center radial function and replacing the entire representation and training object. As this, this goes beyond a simple routine substitution.”
This argument made with respect to the prior art rejection is respectfully inconsistent with Applicant’s arguments made with respect to the written description rejection. If one of ordinary skill in the art could not modify the neural network of Xu to use an elliptic Fourier descriptor, then how could knowledge of the Xu or FourierNet (which, as explained above, uses a Fourier descriptor very similar to Xu’s) prior art explain to one of ordinary skill in the art how the inventor intended to use a trained neural network to estimate an elliptic Fourier descriptor, or make the details of training and inference of such a network “routine and widely known”? If such a substitution is not “simple” or “routine”, then how could the disclosure of FourierNet demonstrate possession of the claimed neural network?
Applicant traverses the previous enablement rejections under 35 U.S.C. 112(a) (Remarks: Pages 9-13). Examiner respectfully disagrees, as explained further below.
Applicant argues that the amendments to the claims have narrowed their scope by reciting a “trained neural network” rather than a more-general “trained machine learning system” (Remarks: Page 10). Examiner agrees that a neural network is more specific than machine learning in general, but also notes that it still covers a wide variety of different architectures. For example, there are over 51,000 U.S. patents classified within CPC G06N 3/02 and its subclasses1, which cover inventions relating to neural networks. In another example, there are over 13,000 U.S. patents classified within CPC 2207/200842, which covers inventions relating to neural networks used for image analysis. Applicant also notes that the amended claims recite more-specific outputs (Remarks: Page 10). Examiner notes that, as explained in the rejection, definition of an output to be obtained from a neural network does not necessarily explain how that neural network obtains that output. Applicant also points to statement #17 in the Declaration (Remarks: Page 10). This statement acknowledges that the claim is broad by not mandating “any particular neural network architecture, training dataset, training regime, or hyperparameter set”, points to prior art including FourierNet and Mask R-CNN, and then asserts that undue experimentation is not required. The statement is respectfully non-persuasive. The breadth of the claim requires more experimentation, not less, because no guidance is given regarding what particular neural network architecture, training dataset, training regime, or hyperparameter set should be used to make and use the invention. Also, as discussed above, Applicant has not shown that any of the noted prior art shows how to make and use the specific trained neural network that estimates the specific elliptic Fourier descriptor required by the claimed invention.
Applicant argues that “the invention is not directed to developing a new neural network architecture” (Remarks: Page 10) and points to a similar statement in the Declaration (#18). These are respectfully non-persuasive for substantially the same reasons discussed above with respect to Page 7 of the Remarks and Statement #12 of the Declaration.
Applicant further argues that disclosures in the prior art, such as Mask R-CNN and FourierNet, are sufficient to enable the claimed invention (Remarks: Pages 11-12; Declaration: Statement 10). As discussed above, while neural networks such as Mask R-CNN and FourierNet were generally known in the prior art, none were capable of estimating an elliptic Fourier descriptor as is specifically required of the trained neural network in the claimed invention. For example, as discussed above, Mask R-CNN estimates a binary mask and none of Applicant’s arguments or declarations has explained how Mask R-CNN could be modified to instead estimate an elliptic Fourier descriptor without undue experimentation. In another example, as also discussed above, Applicant has specifically argued that replacing a single center radial function, as used in FourierNet, with the claimed elliptic Fourier descriptor “goes beyond a simple routine substitution” (Remarks: Page 16, in reference to Xu, which – as explained above – uses a representation very similar to FourierNet), which plainly indicates that undue experimentation would be required.
Applicant next points to Statement #19 in the Declaration, which notes description of the output of the neural network and asserts that selection of a CNN architecture and training would be routine for POSITA (Remarks: Page 12). Examiner respectfully disagrees for substantially the same reasons presented above. I.e., description of the output of a neural network does not explain how to make and use a neural network that obtains that output, and none of the evidence presented by Applicant demonstrates that making and using a neural network that estimates the specific elliptic Fourier descriptor required in the claimed invention was so well-known in the prior art that one of ordinary skill in the art could do it without undue experimentation.
Applicant next points to Statement #20 in the Declaration, which asserts that “implementation choices (e.g., neural-network depth, layer types, optimizer selection, and other hyperparameters) are routine engineering decisions” that do not need to be described in order to provide a working example (Remarks: Page 13). Applicant’s assertions are respectfully non-persuasive. For example, consider the ‘Benbarka’ (“FourierNet: Compact mask representation for instance segmentation using differentiable shape decoders,” 2020) reference. Benbarka describes FourierNet, which is the type prior art that Applicant has asserted one of ordinary skill in the art would be familiar with. Benbarka does not simply describe “the construction of ground truth training data, a precise mathematical definition of the target representation that the neural network is trained to output, and example outputs” (Declaration: Statement #20) and leave the “implementation choices” to the reader’s imagination. Instead, unlike the specification, Benbarka does provide a detailed description of neural-network depth (Sec. 3.2 and Fig. 2, which illustrates the depth of each specific sub-component of the neural network), layer types (e.g., Sec. 3.2 and Fig. 2, backbone, feature pyramid, classification, and shape decoding layers), optimizer selection (Sec. 4, “Stochastic gradient descent (SGD) with momentum (0.9) and weight decay (0.0001) was used for optimization”), and other hyperparameters (e.g., Sec. 3.2, the number and reduction levels of the spatial resolutions; e.g., Sec. 4, “The networks were trained for 12 epochs with an initial learning rate of 0.01 and a mini-batch of 4 images. The learning rate was reduced by a factor of 10 at epochs 8 and 11.”). As an academic paper, Benbarka is clearly intended to be read by those of at least ordinary skill in the art, yet it includes all of these details in order to provide a working example and enable the reader to make and use the FourierNet neural network.
Applicant traverses the previous rejection of claim 11 under 35 U.S.C. 112(b) (Remarks: Page 14), which was based on an interpretation of the term “trained machine learning system” under 35 U.S.C. 112(f). Examiner agrees that the rejection has been overcome, although not necessarily for all of the reasons argued by Applicant. Applicant’s amendment has replaced the term “trained machine learning system” with “neural network,” which is sufficiently structural to preclude interpretation under 35 U.S.C. 112(f). As the term is no longer interpreted under 35 U.S.C. 112(f), the previous rejection based on such interpretation is no longer appropriate and is withdrawn.
Applicant traverses the previous prior art rejections, arguing that the ‘Xu’ and ‘Staib’ references do not teach all elements of the amended claims (Remarks: Pages 15-17). Although he does not agree with all of the arguments made by Applicant, Examiner does agree with Applicant’s argument that
“Staib does not teach or suggest replacing Xu’s one-dimensional radial signature f(θ) with a two-dimensional elliptic Fourier descriptor representation of x(t), y(t). Notably, combining Xu with Staib with Xu would require abandoning Xu’s fundamental use of a single center radial function and replacing the entire representation and training object. As this, this goes beyond a simple routine substitution.” (Remarks: Page 16).
Accordingly, Examiner agrees that the amended claims, as best understood in view of the issues of indefiniteness explained below, are not obvious over Xu and Staib. The previous rejections under 35 U.S.C. 102 and 103 are withdrawn.
Admitted Prior Art
In the Office Action dated May 21, 2025, Examiner took Official Notice of facts in the following instance(s):
At Page 28:
“However, Examiner takes Official Notice that it is old and well-known in the art of image analysis to implement an image processing method as a non-transitory computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method. Such computer implementation advantageously allows the image processing method to be performed quickly and efficiently.”
At Page 29:
“However, Examiner takes Official Notice that it is old and well-known in the art of image analysis to implement an image processing method as a non-transitory computer readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method. Such computer implementation advantageously allows the image processing method to be performed quickly and efficiently.”
Regarding Official Notice, MPEP 2144.03(C) includes the following instructions:
“To adequately traverse such a finding, an applicant must specifically point out the supposed errors in the examiner’s action, which would include stating why the noticed fact is not considered to be common knowledge or well-known in the art.”
“A general allegation that the claims define a patentable invention without any reference to the examiner’s assertion of official notice would be inadequate.”
“If applicant does not traverse the examiner’s assertion of official notice or applicant’s traverse is not adequate, the examiner should clearly indicate in the next Office action that the common knowledge or well-known in the art statement is taken to be admitted prior art because applicant either failed to traverse the examiner’s assertion of official notice or that the traverse was inadequate. If the traverse was inadequate, the examiner should include an explanation as to why it was inadequate.”
In the reply November 21, 2025, Applicant generally alleges that the claims define a patentable invention without any reference to Examiner’s assertion of Official Notice, which is an inadequate traverse.
Therefore, as required by the MPEP, Examiner clearly indicates that the Official Notice statement(s) noted above is/are taken to be admitted prior art because Applicant either failed to traverse it/them or inadequately traversed it/them.
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 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 described in the specification 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 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 described in the specification 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 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 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 function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitation(s) is/are:
the “data processing system …” of claim 11.
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend 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 avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1-3, 6-8, 10-11, and 13 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim 1 recites “inputting the image to a trained neural network, and estimating, by the trained neural network, a representation of a segmentation contour of an object in the image” (reference characters omitted). Additional limitations of claim 1 further define the contour and the estimated representation of the contour.
The estimating limitation is functional at least because it describes a function of estimating a representation of a segmentation contour of an object in an image. The estimating limitation is also computer-implemented at least because it requires a “neural network”, which is computer-implemented.
MPEP 2161.01, Subsection I, provides instructions for determining whether there is adequate written description for a computer-implemented functional claim limitation, including the following:
“Similarly, original claims may lack written description when the claims define the invention in functional language specifying a desired result but the specification does not sufficiently describe how the function is performed or the result is achieved. For software, this can occur when the algorithm or steps/procedure for performing the computer function are not explained at all or are not explained in sufficient detail (simply restating the function recited in the claim is not necessarily sufficient). In other words, the algorithm or steps/procedure taken to perform the function must be described with sufficient detail so that one of ordinary skill in the art would understand how the inventor intended the function to be performed. See MPEP §§ 2163.02 and 2181, subsection IV.”
“When examining computer-implemented functional claims, examiners should determine whether the specification discloses the computer and the algorithm (e.g., the necessary steps and/or flowcharts) that perform the claimed function in sufficient detail such that one of ordinary skill in the art can reasonably conclude that the inventor possessed the claimed subject matter at the time of filing. An algorithm is defined, for example, as "a finite sequence of steps for solving a logical or mathematical problem or performing a task." Microsoft Computer Dictionary (5th ed., 2002). Applicant may "express that algorithm in any understandable terms including as a mathematical formula, in prose, or as a flow chart, or in any other manner that provides sufficient structure." Finisar Corp. v. DirecTV Grp., Inc., 523 F.3d 1323, 1340, 86 USPQ2d 1609, 1623 (Fed. Cir. 2008) (internal citation omitted). It is not enough that one skilled in the art could write a program to achieve the claimed function because the specification must explain how the inventor intends to achieve the claimed function to satisfy the written description requirement.”
Examiner has searched the specification for an explanation of how the estimating step is performed and has identified the following sections as most-pertinent. All citations are to the published application – i.e., US 2023/0298181 A1 – unless otherwise noted.
PNG
media_image1.png
200
400
media_image1.png
Greyscale
PNG
media_image2.png
200
400
media_image2.png
Greyscale
PNG
media_image3.png
200
400
media_image3.png
Greyscale
PNG
media_image4.png
200
400
media_image4.png
Greyscale
PNG
media_image5.png
200
400
media_image5.png
Greyscale
PNG
media_image6.png
200
400
media_image6.png
Greyscale
PNG
media_image7.png
200
400
media_image7.png
Greyscale
PNG
media_image8.png
200
400
media_image8.png
Greyscale
As can be seen from the portions reproduced above, the specification generally restates the desired result of estimating a representation of a segmentation contour of an object in the image using a trained machine learning system without providing any explanation of how this is performed. For example, Figs. 1-4 and pars. [0049]-[0050] and [0054] merely state that a neural network estimates the representation. The neural network is treated as a “black box” and there is no explanation of how the neural network makes such an estimation. For example, what is the architecture of the neural network? I.e., what types of layers does it use and how are they arranged? What type of learning (e.g., supervised, unsupervised, adversarial, etc.) was used to train the neural network? What data sources and loss function(s) were used to train the neural network? What sequence of processing steps is performed by the neural network to produce the segmentation contour representation from the input image? None of these details are provided by the specification.
Given this dearth of detail regarding the structure and functioning of the neural network in the specification, one of ordinary skill in the art would not know how the inventor intends to achieve the claimed function of estimating, by a trained machine learning system, a representation of a segmentation contour of an object in an image and therefore could not reasonably conclude that the inventor possessed the claimed subject matter at the time of filing. It would not be enough if one skilled in the art could hypothetically design and train a neural network to achieve the claimed function because the specification must explain how the inventor intends to achieve the claimed function to satisfy the written description requirement.
Therefore, claim 1 fails to comply with the written description requirement of 35 U.S.C. 112(a).
Claim 11 recites substantially similar limitations and therefore also fails to comply with the written description requirement of 35 U.S.C. 112(a) for substantially the same reasons. Claims 2-3, 6-8, 10, and 13 also fail to comply with the written description requirement of 35 U.S.C. 112(a) at least because they depend from and include the limitations of claim 1.
Claim 7 further recites that “a visibility score is generated by the trained neural network” and further fails to comply with the written description requirement of 35 U.S.C. 112(a) for substantially the same reason as claim 1. I.e., given the general lack of detail regarding the neural network described in the specification, one of ordinary skill in the art could not reasonably conclude that the inventor possessed an invention producing further visibility score outputs from that neural network as required by claim 7.
Claim 8 depends from claim 7 and therefore also further fails to comply with the written description requirement of 35 U.S.C. 112(a) for substantially the same reason as claim 7.
Claims 1-3, 6-8, 10-11, and 13 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the enablement requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to enable one skilled in the art to which it pertains, or with which it is most nearly connected, to make and/or use the invention.
To satisfy the enablement requirement of 35 U.S.C. 112(a), the specification must teach those skilled in the art how to make and use the full scope of the claimed invention without “undue experimentation.”
Claim 1 requires, among other elements, estimating, by a trained neural network, a representation of a segmentation contour of an object in an image. In order to make the invention, one of ordinary skill in the art would have to train a neural network so that it can estimate a representation of a segmentation contour of an object in an image (i.e., perform training). In order to use the invention, one of ordinary skill in the art would have to operate the trained neural network on a given input image (i.e., perform inference).
As discussed above with respect to the written description requirement of 35 U.S.C. 112(a), the specification generally discloses a neural network that performs the estimation, but generally treats the neural network as a “black box” and does not disclose details about the network’s structure, training, or operation. This lack of detail would leave one of ordinary skill in the art with several questions that would have to be answered by experimentation in order to make and use the trained machine learning system required by the claimed invention. For example, what architecture neural network should be used? I.e., what types of layers should be included in the neural network and how should they be arranged? The answers to these questions would affect both training and inference. What type of learning (e.g., supervised, unsupervised, adversarial, etc.) should be used to train the neural network? And what data sources and loss function(s) should be used to train the neural network? The answers to these questions would primarily affect training.
As explained in MPEP 2164.01(a), factors to be considered when assessing whether any necessary experimentation required by the specification is “reasonable” or is “undue” include:
(A) The breadth of the claims;
(B) The nature of the invention;
(C) The state of the prior art;
(D) The level of one of ordinary skill;
(E) The level of predictability in the art;
(F) The amount of direction provided by the inventor;
(G) The existence of working examples; and
(H) The quantity of experimentation needed to make or use the invention based on the content of the disclosure.
Regarding factor (A) The breadth of the claims, the claims cover using any type of neural network in any way that achieves the desired outcome of estimating a representation of a segmentation contour. As discussed above, extensive experimentation would be needed to select an architecture, a training type, a training dataset, and loss functions to make and use a neural network that performs the claimed representation estimation. This factor weighs toward a need for undue experimentation to make and use the claimed invention.
Regarding factors (B) The nature of the invention and (C) The state of the prior art, the invention is focused on the instance segmentation problem recognized within the art of image analysis. See, e.g., par. [0002] of the published specification. As demonstrated by the ‘Hafiz’ (“A survey on instance segmentation: state of the art,” 3 July 2020) reference, many different approaches for performing instance segmentation using a trained machine learning system were known in the prior art. For example, Figures 2-5 and 7-12 of Hafiz illustrate many different neural network architectures used for instance segmentation. In another example, Section 4 lists multiple different datasets used to train machine learning for instance segmentation, each dataset containing images of different types of objects/scenes, under different conditions (e.g., weather) and captured using different types of devices. The high variety of different architectures and datasets shown in Hafiz indicates that a large amount of experimentation would be required to select an appropriate architecture and training dataset for making and using the invention as claimed. These factors weigh toward a need for undue experimentation to make and use the claimed invention.
Furthermore, none of the various prior art examples discussed by Hafiz performs instance segmentation by estimating a representation of a segmentation contour in the manner required by the claims. The techniques taught by Hafiz generally include mask-based techniques, sliding window techniques, region-based techniques, and pixel labeling and clustering techniques (Sec. 2), none of which includes predicting a representation of a segmentation contour as claimed. I.e., none of the cited techniques predicts a Fourier descriptor. Additionally, ‘Benbarka’ (“FourierNet: Compact mask representation for instance segmentation using differentiable shape decoders,” 2020) and ‘Xu’ (“Explicit Shape Encoding for Real-Time Instance Segmentation,” 2019) are examples of prior art that describe estimating Fourier descriptors, but neither estimates elliptic Fourier descriptors. Instead, both estimate Fourier descriptors from a 1D signal of angularly-spaced radius measurements (Benbarka: Sec. 3.1; Xu: e.g., Sec. 3.3.1 and Sec. 3.2.3, Comparison with Other Fitting Methods). As Applicant has acknowledged, substituting elliptical Fourier descriptors for the non-elliptical 1D Fourier descriptors in Xu (and, similarly, Benbarka) would not be simple or routine (Remarks filed November 21, 2025: Page 16). This evidence demonstrates that, as a departure from techniques typically used in the prior art, a relatively large amount of direction or guidance would be needed for one of ordinary skill in the art to make and use a trained neural network that estimates a representation of a segmentation contour as required by the claimed invention. See MPEP 2164.05(a). This further causes factors (B) and (C) to weigh toward a need for undue experimentation to make and use the claimed invention.
This is also pertinent to factors (D) The level of one of ordinary skill, (E) The level of predictability in the art, and (F) The amount of direction provided by the inventor. Because the claimed invention can be seen as a departure from the kinds of techniques known in the prior art as demonstrated by Hafiz, Xu, and Benbarka, one of skill in the art would have less knowledge regarding making and using the type of trained neural network required by the claimed invention. The level of predictability would also be lower because one skilled in the art would have fewer disclosed or known results from which to extrapolate to the claimed invention. One of ordinary skill in the art would also generally need a relatively high amount of direction from the inventor because the claimed invention differs from the types of techniques known in the prior art as demonstrated by Hafiz, Xu, and Benbarka. See MPEP 2164.03. As explained above, the specification provides very little direction and leaves several open questions that would have to be answered by experimentation in order to make and use the claimed invention. Accordingly, these factors weigh toward a need for undue experimentation to make and use the claimed invention.
Regarding factor (G) The existence of working examples, the specification does provide examples of some outputs at Figs. 5-9, but does not provide any example of the actual trained neural network or how it is used to estimate a representation of a segmentation contour. Giving an example of an output of a system, but not describing that system or how it works does not significantly reduce the amount of experimentation needed to make and use a system that can provide such outputs. This factor weighs toward a need for undue experimentation to make and use the claimed invention.
Regarding factor (H) The quantity of experimentation needed to make or use the invention based on the content of the disclosure, as discussed above, the content of the disclosure lacks many of the details needed to actually make and use a trained neural network according to the claimed invention. Many experiments would be required to resolve the many questions left unanswered by the specification. This factor weighs toward a need for undue experimentation to make and use the claimed invention.
Having considered the above factors, it is clear that undue experimentation would be required for one of ordinary skill in the art to make and/or use the claimed invention. Therefore, claim 1 fails to comply with the enablement requirement of 35 U.S.C. 112(a).
Claim 11 recites substantially similar limitations and therefore also fails to comply with the enablement requirement of 35 U.S.C. 112(a) for substantially the same reasons. Claims 2-3, 6-8, 10, and 13 also fail to comply with the enablement requirement of 35 U.S.C. 112(a) at least because they depend from and include the limitations of claim 1.
Claim 7 further recites that “a visibility score is generated by the trained neural network” and further fails to comply with the enablement requirement of 35 U.S.C. 112(a) for substantially the same reason as claim 1. I.e., given the general lack of detail regarding the neural network described in the specification, one of ordinary skill in the art could not make or use a trained neural network producing further visibility score outputs from that neural network as required by claim 7 without undue experimentation.
Claim 8 depends from claim 7 and therefore also further fails to comply with the enablement requirement of 35 U.S.C. 112(a) for substantially the same reason as claim 7.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim(s) 1-3, 6-8, 10, and 13 is/are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites “estimating … a representation of a segmentation contour …, wherein the representation of the segmentation contour is an elliptic Fourier descriptor” (emphasis added; reference characters omitted). Claim 1 further recites that “the estimated representation comprises at least one parameter of a geometric transformation … and a representation of the reference contour” (emphasis added).
These limitations appear to be mutually exclusive, which makes the scope of the claim unclear. Specifically, an elliptic Fourier descriptor does not comprise at least one parameter of a geometric transformation. Indeed, the specification clearly describes the geometric parameter as being different and distinct from the (elliptic) Fourier descriptor – see, e.g., par. [0050] of the published specification and Fig. 3, elements 30’ and 34. It is unclear how the estimated representation can both be an elliptic Fourier descriptor and comprise a geometric transformation parameter, as required by the amended claims. This ambiguity makes the scope of the claim unclear and renders the claim indefinite.
Claim 1 recites “an elliptic Fourier descriptor derived from a Fourier transform of a reference contour.” The elliptic Fourier descriptor is apparently a representation of the reference contour. Claim 1 goes on to recite “a representation of the reference contour.” It is unclear if this later-recited representation is the same as, or different from, the elliptic Fourier descriptor. On the one hand, the claim uses the article “a”, which suggests that a distinct element is being introduced into the claim. On the other hand, as described above, the elliptic Fourier descriptor is a representation of the reference contour, which suggests that they may be the same. This ambiguity makes the scope of the claim unclear and renders the claim indefinite.
Claims 2-3, 6-8, 10, and 13 are also indefinite because they include the indefinite limitations of claim 1.
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.
Claim 11 is rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The scope of the claim includes embodiments that do not fall within at least one of the four categories of patent eligible subject matter.
Claim 11 recites a “data processing system” that performs functions to “input the image to be segmented to the trained neural network, and to reconstruct the segmentation contour of the object from the estimated representation of the segmentation contour by applying an inverse Fourier transform on the elliptic Fourier descriptor” (reference characters omitted). As explained above in Claim Interpretation, the data processing system is being interpreted under 35 U.S.C. 112(f). The corresponding structure, material, or acts in the specification includes a computer program – see, e.g., par. [0071] of the published application.
Claim 11 has been amended to recite that the data processing system comprises a trained neural network. A neural network may be embodied as a computer program. For example, a computer program defining values of weights and rules for applying them to an input to obtain an output.
No structural elements are recited in the claim. Accordingly, the claim covers embodiments that are computer programs per se.
A “computer program per se (often referred to as ‘software per se’) when claimed as a product without any structural recitations” is an example of an embodiment that is “not directed to any of the statutory categories” under 35 U.S.C. 101. MPEP 2106.03, Subsection I. Also, a “claim whose BRI covers both statutory and non-statutory embodiments embraces subject matter that is not eligible for patent protection and therefore is directed to non-statutory subject matter.” MPEP 2106.03, Subsection II.
Therefore, claim 11 is not patent-eligible under 35 U.S.C. 101 because its scope covers embodiments of a computer program per se.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GEOFFREY E SUMMERS whose telephone number is (571)272-9915. The examiner can normally be reached Monday-Friday, 7:00 AM to 3:30 PM ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Chan Park can be reached at (571) 272-7409. 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.
/GEOFFREY E SUMMERS/Examiner, Art Unit 2669
1 See query L96 in attached search history
2 See query L97 in attached search history