DETAILED ACTION
Claims 15-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
Claims 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim(s) 1,2,3,6,7 and 8,9,10,13,14 and 15,16,17 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated1 by Chen et al. (A high-quality self-supervised image denoising method based on SDDW-GAN and CHRNet):
Claim(s) 4,5 and 11 and 18,19,20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al. (A high-quality self-supervised image denoising method based on SDDW-GAN and CHRNet) as applied in claims 1,2,3,6,7 and 8,9,10,13,14 and 15,16,17 above in view of Abdelkader et al. (HEADLESS HORSEMAN: ADVERSARIAL ATTACKS ON TRANSFER LEARNING MODELS:
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al. (A high-quality self-supervised image denoising method based on SDDW-GAN and CHRNet) as applied in claims 1,2,3,6,7 and 8,9,10,13,14 and 15,16,17 above in view of Abdelkader et al. (HEADLESS HORSEMAN: ADVERSARIAL ATTACKS ON TRANSFER LEARNING MODELS) as applied in claims 4,5 and 11 and 18,19,20 further in view of Liu et al. (CAT: Collaborative Adversarial Training):
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 15-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because claim 15 includes non-statutory waves2 to one of skill in the art via claim 15’s “A non-transitory computer-readable medium” is equal to “transmission”3 via applicant’s disclosure’s [0095]:
-- [0095] Further, upon reaching various computer system components, program code means in the form of computer-executable instructions or data structures can be transferred automatically from transmission media to non-transitory computer-readable storage media (devices) (or vice versa). For example, computer-executable instructions or data structures received over a network or data link can be buffered in RAM within a network interface module (e.g., a “NIC”), and then eventually transferred to computer system RAM and/or to less volatile computer storage media (devices) at a computer system. Thus, it should be understood that non-transitory computer-readable storage media (devices) can be included in computer system components that also (or even primarily) utilize transmission media.—
wherein include is defined: to contain, as a whole does parts or any part or element, wherein contain is defined: to be equal to. (Dictionary.com).
Claims 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
PNG
media_image1.png
730
124
media_image1.png
Greyscale
step 0: establish broadest reasonable interpretation as shown in the footnotes of this Office action;
Step 1: claim 1 a process; claim 8 a machine; claim 15 is not statutory;
Step 2A, prong 1:
The claim(s) recite(s) as abstract idea via claim 8 (representative of claims 1 and 15):
training…discrimination…predicted…discrimination…predicted:
8. A system comprising:
a memory component; and
a processing device coupled to the memory component, the processing device to perform operations comprising:
accessing a first training dataset comprising digital images with synthetic noise and corresponding ground-truth digital images for the digital images with synthetic noise;
accessing a second training dataset comprising digital images with natural noise;
training a
generating
generatingdiscrimination between predicted
generatingdiscrimination between predicted
modifying parameters4 of the domain gap generative adversarial network based on the first discrimination
1. A computer-implemented method comprising:
accessing a first training dataset comprising digital images with synthetic noise and corresponding ground-truth digital images for the digital images with synthetic noise;
accessing a second training dataset comprising digital images with natural noise; training a domain gap generative adversarial network by:
generating, utilizing the domain gap generative adversarial network, predicted denoised images from the digital images with synthetic noise and predicted denoised images from the digital images with natural noise;
utilizing a discriminator to generate a first discrimination between the ground-truth digital images for the digital images with synthetic noise and the predicted denoised images for the digital images with synthetic noise, and a second discrimination between the predicted denoised images for the digital images with natural noise and the predicted denoised images for the digital images with synthetic noise; and
modifying parameters of the domain gap generative adversarial network based on the first discrimination and the second discrimination.
15. A non-transitory computer-readable medium5 storing executable instructions which, when executed by a processing device, cause the processing device to perform operations comprising:
accessing a first training dataset comprising digital images with synthetic noise and corresponding ground-truth digital images for the digital images with synthetic noise;
accessing a second training dataset comprising digital images with natural noise;
training a domain gap generative adversarial network by:
generating, utilizing the domain gap generative adversarial network, predicted denoised images from the digital images with synthetic noise and predicted denoised images from the digital images with natural noise;
utilizing a discriminator to generate a first discrimination between the ground-truth digital images for the digital images with synthetic noise and the predicted denoised images for the digital images with synthetic noise, and a second discrimination between the predicted denoised images for the digital images with natural noise and the predicted denoised images for the digital images with synthetic noise; and
modifying parameters of the domain gap generative adversarial network based on the first discrimination and the second discrimination.
Step 2A, prong 2:
This judicial exception is not integrated into a practical application because the additional elements is not improving the functioning of a computer in view of applicant’s disclosure at [0023]:
--[0023] As suggested above, embodiments of the domain gap generative adversarial system provide certain improvements or advantages over conventional systems. More specifically, the training pipeline of the domain gap generative adversarial system bridges the domain gap between natural and synthetic noise. To illustrate, the domain gap generative adversarial system improves accuracy relative to conventional systems by generating clearer denoised images relative to conventional generative adversarial networks. For example, the domain gap generative adversarial system trains a neural network with improved accuracy over those trained using only digital images with synthetic noise. Specifically, by utilizing both a paired ground-truth dataset of digital images with synthetic noise and an unpaired dataset with natural noise, the domain gap generative adversarial system trains the domain gap generative adversarial network to more accurately process digital images with natural noise. Accordingly, the domain gap generative adversarial system improves the clarity of denoised images relative to conventional systems.—
PNG
media_image2.png
597
971
media_image2.png
Greyscale
Step 2B:
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements (claim 8’s computer stuff) with the abstract (train, discriminate, predict) adheres to the conventional in view of applicant’s disclosure and 35 USC 112(a)6:
PNG
media_image3.png
1390
920
media_image3.png
Greyscale
PNG
media_image4.png
384
877
media_image4.png
Greyscale
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1,2,3,6,7 and 8,9,10,13,14 and 15,16,17 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated7 by Chen et al. (A high-quality self-supervised image denoising method based on SDDW-GAN and CHRNet):
PNG
media_image5.png
730
255
media_image5.png
Greyscale
Re 1., Chen discloses A computer-implemented method (likewise) comprising8:
accessing a first training dataset comprising digital images with synthetic9 noise (or likewise “The RENOIR dataset uses three cameras to…Suppose there is a noise distribution called A”, pg. 3, 3.1 Data acquisition, 1st S & 3.2 Proposed network, 2nd para, 2nd S: fig. 1: feigned checker-board noise n(j)) 10
PNG
media_image6.png
633
1047
media_image6.png
Greyscale
;
accessing a second training dataset comprising digital images with natural noise (or likewise a subset via “The RENOIR dataset…in-cludes two noise images”, pg. 3, 3.1 Data acquisition, 2nd S);
training a 11 generative adversarial network (or likewise “trains the GAN”), pg. 4, lcol, last para, penult S) by:
generating12 13 images from the digital images with synthetic noise (or likewise “to represent noisy images…the network predicts the output to get Y”, pg. 2, rcol, bullet (2)b. & pg. 4, lcol, 3rd para, 1st S)
PNG
media_image7.png
605
1124
media_image7.png
Greyscale
;
utilizing a discriminator (or likewise circled “dual discriminators”, pg. 4, lcol, last para, 2nd S) to generate a first discrimination between
PNG
media_image8.png
633
1047
media_image8.png
Greyscale
; and
modifying parameters of the domain gap generative adversarial network based on the first discrimination (or likewise “parameters…adapated…given …denoised images”, pg. 9 rcol, 2nd S) 14.
Re 2., Chen discloses The computer-implemented method of claim 1, further comprising:
determining image loss (pg. 4 equation (1))
modifying the parameters of the domain gap generative adversarial network further based on the image loss
Re 3., Chen discloses The computer-implemented method of claim 1, further comprising:
utilizing the first discrimination
modifying the parameters of the
Re 6., Chen discloses The computer-implemented method of claim 1, further comprising generating4.5 Real-world noise experiments, 1st para, 2nd S: fig. 6:
PNG
media_image9.png
408
967
media_image9.png
Greyscale
).
Re 7., Chen discloses The computer-implemented method of claim 1, further comprising generating the first training dataset by adding synthetic noise (“additive noise is given”, pg. 2, rcol, bullet (3), 1st S) to a set of digital images (via fig. 1: two plus “+” signs:
PNG
media_image10.png
502
955
media_image10.png
Greyscale
.
Claim 8 is rejected like claim 1:
Re 8., Chen discloses A system comprising:
a memory component (or likewise “64 GB RAM”, pg. 7, 4.1 Experimental environment and settings, 1st para); and
a processing device coupled to the memory component, the processing device to perform operations comprising:
accessing a first training dataset comprising digital images with synthetic noise
accessing a second training dataset comprising digital images with natural noise;
training a domain gap generative adversarial network by:
generating
generating
generating
modifying parameters of the domain gap generative adversarial network based on the first discrimination and the second discrimination.
Claim 9 is rejected like claim 2:
Re 9., Chen discloses The system of claim 8, wherein the operations further comprise:
determining image loss
modifying the parameters of the domain gap generative adversarial network further based on the image loss
Claim 10 is rejected like claim 3:
Re 10. The system of claim 8, wherein the operations further comprise:
utilizing the first discrimination
modifying the parameters of the
Claim 13 is rejected like claim 6:
Re 13., Chen discloses The system of claim 8, wherein the operations further comprise generating
Claim 14 is rejected like claim 7:
Re 14.m Chen discloses The system of claim 8, wherein the operations further comprise generating the first training dataset by adding synthetic noise to a set of digital images.
Claim 15 is rejected like claims 1 and 8:
Re 15., Chen discloses A non-transitory computer-readable medium storing executable instructions which, when executed by a processing device, cause the processing device to perform operations comprising:
accessing a first training dataset comprising digital images with synthetic noise
accessing a second training dataset comprising digital images with natural noise;
training a domain gap generative adversarial network by:
generating
utilizing a discriminator to generate a first discrimination between
modifying parameters of the domain gap generative adversarial network based on the first discrimination
Claim 16 is rejected like claims 2 and 9:
Re 16., Chen discloses The non-transitory computer-readable medium of claim 15, wherein the operations further comprise:
determining image loss and
modifying the parameters of the
Claim 17 is rejected like claims 3 and 10:
Re 17., Chen discloses The non-transitory computer-readable medium of claim 15, wherein the operations further comprise:
utilizing the first discrimination
modifying the parameters of the .
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) 4,5 and 11 and 18,19,20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al. (A high-quality self-supervised image denoising method based on SDDW-GAN and CHRNet) as applied in claims 1,2,3,6,7 and 8,9,10,13,14 and 15,16,17 above in view of Abdelkader et al. (HEADLESS HORSEMAN: ADVERSARIAL ATTACKS ON TRANSFER LEARNING MODELS:
PNG
media_image11.png
730
471
media_image11.png
Greyscale
Re 4., Chen teaches The computer-implemented method15 of claim 1,16 further comprising17:
utilizing18 the first discrimination to19 determine ground-truth (or likewise “ground truth”, pg. 3, rcol, 3.1 Data acquisition, 1st para, 2nd S) logits based on
utilizing20 21 determine ground-truth (or said likewise “ground truth”, pg. 3, rcol, 3.1 Data acquisition, 1st para, 2nd S) logits based on
applying22 a generative adversarial network 23 the ground-truth logits and the ground-truth logits to24 determine generative adversarial network loss (pg. 4 equation (1)).
Chen does not teach the difference of claim 4 of:
(ground-truth)25 26 logits27…28
(ground-truth) logits…
the (ground-truth) logits and the (ground-truth) logits.
PNG
media_image2.png
597
971
media_image2.png
Greyscale
Abdelkader teach the difference of claim 4 of:
(ground-truth) th highest logit”, 3rd page, Algorithm 1: Headless Horseman Algorithm, 8th line)…
(ground-truth) logits (or likewise “ith highest logit”, 3rd page, Algorithm 1: Headless Horseman Algorithm, 8th line)…
the (ground-truth) logits th highest logit”, 3rd page, Algorithm 1: Headless Horseman Algorithm, 8th line).
Since Chen suggests using clean image for classification and cites to others, 1st page, lcol, 1st S:
1. Introduction
Image denoising has been an essential part of the computer vision field, while separating noise from noisy images and retaining clean images is an important preprocessing step in many vision tasks, such as image classification (Ning, Tian, Yu, Li, Bai, & Wang, 2022), image segmentation (Yu et al., 2023), medical images (Wang, Li, Du, Xiao, & Gao, 2022), image compression (Ranjan, & Kumar, 2022), video
denoising (Zhang and Zhou, 2023), target detection (Ni, Luo, Wang, Liang, & Zhang, 2023), and remote sensing images (He, Gong, Hu, & Li, 2022). According to the degradation model Y = X + V, the goal of image denoising is to recover a noise-free image X from a noisy observation Y by reducing the noise V. People are inevitably affected by imaging equipment and external environmental noise, in the process of image acquisition, digitization, and transmission, which leads to inaccurate or even incorrect acquired information and thus reduces the visual effect of the image. For example, cell phone photos on specific occasions (e.g., at night) are usually affected by unknown noise. Removing these noises is vital to improve the visual experience of users.
one of skill in the art of classification would or could have done is look to others for image classification using clean images and thus make Chen’s be as Adbelkader’s seeing in the change that “Transfer learning facilitates29 the training of task-specific classifiers using pre-trained models as feature extractors.”, Abdelkader, abstract, 1st S, via explicit creative or even routine steps a)-f):
a) create a bunch of clean images “Y” from the “Noisy Image Dataset” using Chen’s fig. 1:
a1) using RENOIR image dataset:
a1.1) create a subset of noise images and a subset of ground truth images of a dark scene of flowers/roses indoors:
PNG
media_image12.png
653
1105
media_image12.png
Greyscale
b) obtain a pre-trained classifier logits code from Abdelkader’s “PyTorch”-“code repository”, page 3, section 4 EXPERIMENTS, 1st para, last S:
PNG
media_image13.png
393
847
media_image13.png
Greyscale
b1) check for security sensitivity of the classifier (i.e., a pre-trained neural network with possible viruses/attacks downloaded off the public internet via “synthetic prediction logits”, Abdelkader, 2nd page, 3.1 Centroid-based attack, 1st para, 3rd S);
c) install the PyTorch logits pre-trained code free of viruses/attacks in CHEN’s 64 GB (Giga-Byte) RAM;
d) run the virus-free PuTorch logits pre-trained code in the 64 GB RAM;
e) input said CHEN’s clean image bunch Y and corresponding ground-truth images into the 64 GB RAM installed pre-trained logits PyTorch classifier code:
e1) re-train the pre-trained (cat) neural-network classifier, i.e., transfer-learning from whatever images (cats) to roses/flowers, on flowers/roses images indoors; and
f) see what happens (easier “ground-truth” “logits” (cat) classification of dark-rose-scene RENOIR data subset noisy images Y in relation to a centroid ground truth, Abdelkader, section 4.1 Centroid-base attacks):
PNG
media_image14.png
632
561
media_image14.png
Greyscale
PNG
media_image15.png
1416
1025
media_image15.png
Greyscale
Re 5., Chen of the combination of Chen,Abdelkader teaches The computer-implemented method of claim 4, further comprising:
utilizing a first discriminator (or said likewise circled “dual discriminators”, pg. 4, lcol, last para, 2nd S) to generate the first discrimination between
utilizing a second discriminator (or said likewise circled “dual discriminators”, pg. 4, lcol, last para, 2nd S) 30 between
Claim 11 is rejected like claim 4:
Re 11., Chen of the combination of Chen,Abdelkader teaches The system of claim 8, wherein the operations further comprise:
utilizing the first discriminator to determine ground-truth logits based on the ground-truth digital images for the digital images with synthetic noise;
utilizing the first discriminator to determine ground-truth logits based on the predicted
applying a generative adversarial network the ground-truth logits and the ground-truth logits to determine generative adversarial network loss.
Claim 18 is rejected like claims 4 and 11:
Re 18., Chen of the combination of Chen,Abdelkader teaches The non-transitory computer-readable medium of claim 15, wherein the operations further comprise:
utilizing the first discrimination to determine ground-truth logits based on
utilizing the second discrimination to determine ground-truth logits
applying a generative adversarial network logits and the ground-truth logits to determine generative adversarial network loss.
Claim 19 is rejected like claim 5:
Re 19., Chen of the combination of Chen,Abdelkader teaches The non-transitory computer-readable medium of claim 18, wherein the operations further comprise:
utilizing a first discriminator to generate the first discrimination between
utilizing a second discriminator
Claim 20 is rejected like claim 6:
Re 20., Chen of the combination of Chen,Abdelkader teaches The non-transitory computer-readable medium of claim 18, wherein the operations further comprise generating.
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al. (A high-quality self-supervised image denoising method based on SDDW-GAN and CHRNet) as applied in claims 1,2,3,6,7 and 8,9,10,13,14 and 15,16,17 above in view of Abdelkader et al. (HEADLESS HORSEMAN: ADVERSARIAL ATTACKS ON TRANSFER LEARNING MODELS) as applied in claims 4,5 and 11 and 18,19,20 further in view of Liu et al. (CAT: Collaborative Adversarial Training):
PNG
media_image16.png
730
501
media_image16.png
Greyscale
Re 12., Abdelkader of the combination of Chen,Abdelkader teaches The system of claim 11, wherein the operations further comprise:
utilizing the second discriminator to determine natural logits based on the predicted digital images for the digital images with synthetic noise;
utilizing the second discriminator to determine synthetic logits (or likewise “synthetic prediction logits”, Abdelkader, 2nd page, 3.1 Centroid-based attack, 1st para, 3rd S) based on the predicted
applying the generative adversarial network natural logits to31 determine the generative adversarial network loss.
Abdelkader of the combination of Chen,Abdelkader does not teach the difference of claim 12 of:
natural (logits)32.
Liu teach the difference of claim 12 of:
natural (logits)33 (or likewise “natural logits”, pg. 3, rcol, penult S).
Since Abdelkader suggests that classifying images is widely practiced with corresponding problems thereof, page 1, 1. INTRODUCTION, 1st two paragraphs:
Neural networks are powerful tools for solving computer vi sion problems, but training them from scratch requires huge amounts of data and compute time [1, 2]. One of the most popular frameworks for reducing data requirements is transfer learning [3] in which a pre-trained network (usually trained on a large labeled dataset like ImageNet) is used to extract low dimensional features from images. A new linear classifier head is then trained to classify images using a small task-specific dataset and the corresponding number of outputs. Transfer learning is widely used in practice thanks to the availability of standard pre-trained models within common deep learning frameworks like PyTorch and TensorFlow.
In this paper, we study the security vulnerabilities intro duced by simple transfer learning strategies. Deep neural networks are known to be vulnerable to adversarial attacks: inputs that have been maliciously crafted to fool a victim network. These attacks generally follow two paradigms: (i) white-box attacks, in which the attacker has complete knowl edge of the victim network (both the network architecture and weights) and (ii) black-box attacks in which the attacker does not know the victim network but can query its output label on chosen inputs [4, 5].
one of skill in the art of classifiers could or would have done is refer to other teachings of classification as a solution to the classification problem and thus make Abdelkader’s of the combination of Chen,Abdelkader be as Liu’s seeing in the change “both networks learn better decision boundaries than learning alone to obtain better adversarial robustness.”, Liu, page 4; rcol, 1st para, last S, via additional explicit creative or even routine steps: e2:
a) create a bunch of clean images “Y” from the “Noisy Image Dataset” using Chen’s fig. 1:
a1) using RENOIR image dataset:
a1.1) create a subset of noise images and a subset of ground truth images of a dark scene of flowers/roses indoors:
PNG
media_image12.png
653
1105
media_image12.png
Greyscale
b) obtain a pre-trained classifier logits code from Abdelkader’s “PyTorch”-“code repository”, page 3, section 4 EXPERIMENTS, 1st para, last S:
PNG
media_image13.png
393
847
media_image13.png
Greyscale
b1) check for security sensitivity of the classifier (i.e., a pre-trained neural network with possible viruses/attacks downloaded off the public internet via “synthetic prediction logits”, Abdelkader, 2nd page, 3.1 Centroid-based attack, 1st para, 3rd S);
c) install the PyTorch logits pre-trained code free of viruses/attacks in CHEN’s 64 GB (Giga-Byte) RAM;
d) run the virus-free PuTorch logits pre-trained code in the 64 GB RAM;
e) input said CHEN’s clean image bunch Y and corresponding ground-truth images into the 64 GB RAM installed pre-trained logits PyTorch classifier code:
e1) re-train the pre-trained (cat) neural-network classifier, i.e., transfer-learning from whatever images (cats) to roses/flowers, on flowers/roses images indoors;
e2) write new-code in the PyTorch classifier code calling an adversarial training framework program based on Liu’s training framework of fig. 2:
PNG
media_image17.png
775
999
media_image17.png
Greyscale
e2.11) input the pre-trained PyTorch classifier as coded “model: f”;
e.2.12) input another pre-trained classier (possibly already virus infected/attacked via a synthetic logit) as coded “model:g”;
e2.2) run the classifier training framework code-program;
e2.3 return back to pre-trained classifier PyTorch code ; and
f) see what happens (easier “ground-truth” “logits” (cat) classification of dark-rose-scene RENOIR data subset noisy images Y in relation to a centroid ground truth, Abdelkader, section 4.1 Centroid-base attacks, with better adversarial decision boundaries):
PNG
media_image14.png
632
561
media_image14.png
Greyscale
PNG
media_image15.png
1416
1025
media_image15.png
Greyscale
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
He et al. (De-Noising of Photoacoustic Microscopy Images by Attentive Generative Adversarial Network)
He probabilistically predicts a “de-noised image”: fig. 1: “Discriminator”, via page 1351, rcol continued in page 1352:
PNG
media_image18.png
566
1054
media_image18.png
Greyscale
2) Discriminator Network: The discriminator illustrated in Fig. 1 is an 8-layer CNN with the number of filters as 64, 64, 128, 128, 256, 256, 512, and 512, respectively. Each of the convolutional layers has a kernel size of 3×3 and is subsequently equipped with a LReLU and a batch normalization (BN). In the end, there are two fully-connected (FC) layers with 1024 outputs and a single output, respectively. The input of the discriminator is the de-noised image or its corresponding clean image (ground truth), and the output is the probability that the de-noised image is true.
as the closest to the claimed “generating
IDS cited Hong et al. (End-to-End Unpaired Image Denoising with Conditional Adversarial Networks)
Hong teaches denoising in the context of probabilistically predicting a generated pseudo-noisy image, page 4143, rcol, last paragraph: fig. 2: “D”:
PNG
media_image19.png
488
1027
media_image19.png
Greyscale
Discriminative Network Architecture The discriminator takes a real noisy image from Y ′ or generated pseudo noisy image from X′ as input and returns the probability that the image is sampled from real noisy data. The architecture of the discriminator is illustrated in Fig. 4 which is similar to DCGAN(Radford, Metz, and Chintala 2015). The difference is that we have removed the batch normalization layers and the last sigmoid layer as WGAN-GP (Gulrajani et al. 2017). In the training process, we use 64 × 64 image patches to train the GAN network. The images fed to the discriminator are X′ and Y ′ that may have different image contents. To make the discriminator focus more on noise, we apply a sharpening technique by subtracting a image from its local-mean filtered version. The sharpened images and original images serve as two group of channels of the input for the discriminator. This sharpening technique is critical to the success of model in performance. The kernel size for calculating local mean is a hyper-parameter and is set to 3 × 3 in our experiments.
as the closest to the claimed “generating
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 MPEP 2131 Anticipation — Application of 35 U.S.C. 102 [R-08.2017], 2nd para, 2nd to last S:
The elements must be arranged as required by the claim, but this is not an ipsissimis verbis test, i.e., identity of terminology is not required. In re Bond, 910 F.2d 831, 15 USPQ2d 1566 (Fed. Cir. 1990).
2 MPEP 2106 II. ESTABLISH BROADEST REASONABLE INTERPRETATION OF CLAIM AS A WHOLE, 2nd para: Claim interpretation affects the evaluation of both criteria for eligibility. For example, in Mentor Graphics v. EVE-USA, Inc., 851 F.3d 1275, 112 USPQ2d 1120 (Fed. Cir. 2017), claim interpretation was crucial to the court’s determination that claims to a "machine-readable medium" were not to a statutory category. In Mentor Graphics, the court interpreted the claims in light of the specification, which expressly defined the medium as encompassing "any data storage device" including random-access memory and carrier waves. Although random-access memory and magnetic tape are statutory media, carrier waves are not because they are signals similar to the transitory, propagating signals held to be non-statutory in Nuijten. 851 F.3d at 1294, 112 USPQ2d at 1133 (citing In re Nuijten, 500 F.3d 1346, 84 USPQ2d 1495 (Fed. Cir. 2007)). Accordingly, because the BRI of the claims covered both subject matter that falls within a statutory category (the random-access memory), as well as subject matter that does not (the carrier waves), the claims as a whole were not to a statutory category and thus failed the first criterion for eligibility.
3 transmission: Radio and Television. the broadcasting of electromagnetic waves from one location to another, as from a transmitter to a receiver. (Dictionary.com)
4 parameter: Computers. a variable that must be given a specific value during the execution of a program or of a procedure within a program. (Dictionary.com)
5 Applicant’s disclosure (“non-transitory computer-readable storage media (devices)” is equal to “other medium” and equal to “utilize transmission media”:
-- [0093] Non-transitory computer-readable storage media (devices) includes RAM, ROM, EEPROM, CD-ROM, solid state drives (“SSDs”) (e.g., based on RAM), Flash memory, phase-change memory (“PCM”), other types of memory, other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store desired program code means in the form of computer-executable instructions or data structures and which can be accessed by a general purpose or special purpose computer. -- wherein include is defined: to contain, as a whole does parts or any part or element, wherein contain is defined: to be equal to. (Dictionary.com)
--[0095] Further, upon reaching various computer system components, program code means in the form of computer-executable instructions or data structures can be transferred automatically from transmission media to non-transitory computer-readable storage media (devices) (or vice versa). For example, computer-executable instructions or data structures received over a network or data link can be buffered in RAM within a network interface module (e.g., a “NIC”), and then eventually transferred to computer system RAM and/or to less volatile computer storage media (devices) at a computer system. Thus, it should be understood that non-transitory computer-readable storage media (devices) can be included in computer system components that also (or even primarily) utilize transmission media.
-- wherein include is defined: to contain, as a whole does parts or any part or element, wherein contain is defined: to be equal to. (Dictionary.com)
6 MPEP 2106.05(d) Well-Understood, Routine, Conventional Activity [R-07.2022]
I. EVALUATING WHETHER THE ADDITIONAL ELEMENTS ARE WELL-UNDERSTOOD, ROUTINE, CONVENTIONAL ACTIVITY
2. A factual determination is required to support a conclusion that an additional element (or combination of additional elements) is well-understood, routine, conventional activity, 2nd para:.
As such, an examiner should determine that an element [claim 8’s “discriminator” that is outside the broadest reasonable interpretation of claim 8] (or combination of elements) is well-understood, routine, conventional activity only when the examiner can readily conclude, based on their expertise in the art, that the element is widely prevalent or in common use in the relevant industry. The analysis as to whether an element (or combination of elements) is widely prevalent or in common use is the same as the analysis under 35 U.S.C. 112(a) as to whether an element (the disclosure’s “zero-sum game”, applicant’s disclosure [0039] 2nd S) is so well-known that it need not be described in detail in the patent specification. See Genetic Techs. Ltd. v. Merial LLC, 818 F.3d 1369, 1377, 118 USPQ2d 1541, 1546 ( Fed. Cir. 2016) (supporting the position that amplification was well-understood, routine, conventional for purposes of subject matter eligibility by observing that the patentee expressly argued during prosecution of the application that amplification was a technique readily practiced by those skilled in the art to overcome the rejection of the claim under 35 U.S.C. 112, first paragraph);
7 MPEP 2131 Anticipation — Application of 35 U.S.C. 102 [R-08.2017], 2nd para, 2nd to last S:
The elements must be arranged as required by the claim, but this is not an ipsissimis verbis test, i.e., identity of terminology is not required. In re Bond, 910 F.2d 831, 15 USPQ2d 1566 (Fed. Cir. 1990).
8 BROAD CLAIM LANGUAGE: -ing (of “comprising”): a suffix of nouns formed from verbs, expressing the action of the verb or its result, product, material, etc. (the art of building; a new building; cotton wadding ).wherein etc is defined: and others; and so forth; and so on (used to indicate that more of the same sort or class might have been mentioned, but for brevity have been omitted). wherein so is defined: likewise or correspondingly; also; too. (Dictionary.com)
9 synthetic: not real or genuine; artificial; feigned. (Dictionary.com)
10 and: (used to connect alternatives). (Dictionary.com)
11 “domain gap” is coordinate adjective that individually coordinates with “generative adversarial” to modify “network” : domain gap network AND/OR generative adversarial network.
12 comma: the punctuation mark(,) indicating a slight pause in the spoken sentence and used where there is a listing of items or to separate a nonrestrictive clause or phrase from a main clause (Dictionary.com)
13 “predicted denoised” are coordinate adjective that independently coordinate to modify “images”
14 The above (as well as the below) crossed-out text “does not limit the scope of” the claims (1) and “the prior art teaches the element if one of the alternatives is taught by the prior art” via MPEP 2143.03 All Claim Limitations Must Be Considered [R-01.2024], 3rd para: As a general matter, the grammar (‘and” and coordinate adjectives) and ordinary meaning of terms as understood by one having ordinary skill in the art used in a claim will dictate whether, and to what extent, the language limits the claim scope. Language that suggests or makes a feature or step optional but does not require that feature or step does not limit the scope of a claim under the broadest reasonable claim interpretation. In addition, when a claim requires selection of an element from a list of alternatives, the prior art teaches the element if one of the alternatives is taught by the prior art. See, e.g., Fresenius USA, Inc. v. Baxter Int’l, Inc., 582 F.3d 1288, 1298, 92 USPQ2d 1163, 1171 (Fed. Cir. 2009), wherein scope is defined: Linguistics, Logic. the range of words or elements of an expression (claim 1) over which a modifier (e.g., a patent examiner) or operator (e.g., me) has control. (Dictionary.com)
15 subject
16 Claim Interpretation: ignore this comma “,”, that causes a sentence fragment in claim 4
17 verb
18 participle (i.e., adjective)
19 to: (used for expressing aim, purpose, or intention). (Dictionary.com)
20 gerund (i.e. noun)
21 to: (used for expressing aim, purpose, or intention). (Dictionary.com)
22 participle (i.e., adjective)
23 to: (used for expressing aim, purpose, or intention). (Dictionary.com)
24 LONG RANGE CLAIM SCOPE: to: preposition: any member of a class of words found in many languages that are used before nouns (“loss”), pronouns, or other substantives to form phrases (“to determine generative adversarial network loss”) functioning as modifiers of verbs, nouns, or adjectives (“applying” is a participle used as an adjective modifying the noun “method” of the subject “computer-implemented method” of claim 4), and that typically express a spatial, temporal, or other relationship, as in, on, by, to, since. (Dictionary.com)
25 (italics) represent claim limitations already taught
26 As discussed in the rejection of claim 1, the above (as well as the below) crossed-out text “does not limit the scope of” the claims (1) and “the prior art teaches the element if one of the alternatives is taught by the prior art” via MPEP 2143.03 All Claim Limitations Must Be Considered [R-01.2024] As a general matter, the grammar (‘and” and coordinate adjectives) and ordinary meaning of terms as understood by one having ordinary skill in the art used in a claim will dictate whether, and to what extent, the language limits the claim scope. Language that suggests or makes a feature or step optional but does not require that feature or step does not limit the scope of a claim under the broadest reasonable claim interpretation. In addition, when a claim requires selection of an element from a list of alternatives, the prior art teaches the element if one of the alternatives is taught by the prior art. See, e.g., Fresenius USA, Inc. v. Baxter Int’l, Inc., 582 F.3d 1288, 1298, 92 USPQ2d 1163, 1171 (Fed. Cir. 2009), wherein scope is defined: Linguistics, Logic. the range of words or elements of an expression (claim 1) over which a modifier (e.g., a patent examiner) or operator (e.g., me) has control. (Dictionary.com)
27 THE CLAIMED INVENTION AS A WHOLE, regarding “logits”:
Applicant’s disclosed problem is:
--[0002] Although conventional systems denoise digital images, such systems have a number of problems in relation to accuracy and efficiency. For instance, conventional systems inaccurately denoise images because they utilize neural networks trained using synthetic noise. Because synthetic noise uniformly places noise on a digital image, synthetic noise inaccurately portrays the noise distribution of the natural noise of digital images. Because of this disparity, or domain gap, between natural noise and synthetic noise, conventional image editing systems generate inaccurate results because their models cannot accurately address natural noise in digital images.--
Chen teaches a similar problem regarding an “inaccurate”, Chen, pg. 1, 1. Introduction, rcol, 2nd S, image
Applicant’s disclosed solution is (see applicant’s fig. 4 (reproduced below): “Ground Truth Digital Image”’s arrow paths):
--[0023] As suggested above, embodiments of the domain gap generative adversarial system provide certain improvements or advantages over conventional systems. More specifically, the training pipeline of the domain gap generative adversarial system bridges the domain gap between natural and synthetic noise. To illustrate, the domain gap generative adversarial system improves accuracy relative to conventional systems by generating clearer denoised images relative to conventional generative adversarial networks. For example, the domain gap generative adversarial system trains a neural network with improved accuracy over those trained using only digital images with synthetic noise. Specifically, by utilizing both a paired ground-truth dataset of digital images with synthetic noise and an unpaired dataset with natural noise, the domain gap generative adversarial system trains the domain gap generative adversarial network to more accurately process digital images with natural noise. Accordingly, the domain gap generative adversarial system improves the clarity of denoised images relative to conventional systems.—
While Chen teaches “ground truth”, pg. 3, rcol., 3.1 Data acquisition, 1st para, 2nd S, Chen does not appear to use the ground truth and instead teaches “visual quality is the main metric”, Chen, pg. 8, 4.5 Real-world noise experiments, 1st para, last S.
Since the disclosed solution is lacking in claims 1 and 4, the difference of claim 4 is an indication of obviousness
28 ellipses (…) represent claim limitations already taught
29 facilitate: to make easier or less difficult; help forward (an action, a process, etc.). (Dictioanry.com)
30 the claimed “to generate the second discrimination” does not limit the scope of claim 5 under the broadest reasonable interpretation via said MPEP 2143.03 All Claim Limitations Must Be Considered [R-01.2024], 3rd para:
As a general matter, the grammar (‘and” and coordinate adjectives) and ordinary meaning of terms as understood by one having ordinary skill in the art used in a claim will dictate whether, and to what extent, the language limits the claim scope. Language that suggests or makes a feature or step optional but does not require that feature or step does not limit the scope of a claim under the broadest reasonable claim interpretation. In addition, when a claim requires selection of an element from a list of alternatives, the prior art teaches the element if one of the alternatives is taught by the prior art. See, e.g., Fresenius USA, Inc. v. Baxter Int’l, Inc., 582 F.3d 1288, 1298, 92 USPQ2d 1163, 1171 (Fed. Cir. 2009), wherein scope is defined: Linguistics, Logic. the range of words or elements of an expression (claim 1) over which a modifier (e.g., a patent examiner) or operator (e.g., me) has control. (Dictionary.com)
31 LONG RANGE CLAIM SCOPE: to: preposition: any member of a class of words found in many languages that are used before nouns (“loss”), pronouns, or other substantives to form phrases (“to determine generative adversarial network loss”) functioning as modifiers of verbs, nouns, or adjectives (“applying” is a participle used as an adjective modifying the noun “method” of the subject “computer-implemented method” of claim 4), and that typically express a spatial, temporal, or other relationship, as in, on, by, to, since. (Dictionary.com)
32 (italics) represent claim limitations already taught
33 (italics) represent claim limitations already taught
Prosecution Insights