DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Applicant’s arguments, see page 9, filed 4/10/2026, with respect to the specification objection have been fully considered and are persuasive. The objection of the specification has been withdrawn.
Applicant’s arguments, see page 9, filed 4/10/2026, with respect to claim objections have been fully considered and are persuasive. The objection of the claims has been withdrawn.
Applicant’s arguments, see page 9, filed 4/10/2026, with respect to 101 rejections have been fully considered and are persuasive. The 101 rejection of the claims has been withdrawn.
Applicant’s arguments, see page 10, filed 4/10/2026, with respect to the rejection(s) of claim(s) 1-5 and 7-13 under 112(b) have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Fang et al. The reference is a conference paper published in 2022 that discloses the claimed invention through the figures 1(a) and (b). In addition, the reference discloses the cross knowledge distillation for image super-resolution in the pages explaining the purpose of the figures. Thus, based on the newly applied reference, the features of the claims are disclosed below.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 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.
Claims 1-4, 7-9 and 11-13 are rejected under 35 U.S.C. 102(a)(1) based upon a public use or sale or other public availability of the invention. Document titled “Cross Knowledge Distillation for Image Super-Resolution (Publish Date: December 2022). ICVIP '22: Proceedings of the 2022 6th International Conference on Video and Image Processing. Hereinafter referred to as “Fang et al”.
Re claim 1: (CURRENTLY AMENDED) Fang et al discloses an image super-resolution method based on a knowledge distillation compression model, comprising:
(1) acquiring a trained large teacher network and a small student network to be trained respectively, wherein the teacher network is an open and pre-trained network; the student network is a network obtained after reducing a depth of the teacher network, which will be used for actual deployment; the teacher network and the student network are respectively divided into two modules according to a network depth to obtain a first teacher module, a second teacher module, a first student module and the a second student module; based on a softened interface which integrates semantic features of a previous layer with features extracted from a current layer, the first teacher module, the softened interface and the second student module are cascaded in sequence, and the first student module, the softened interface and the second teacher module are cascaded, to form two optimized paths for knowledge distillation (e.g. as seen in figure 1(a), the system in the paper shows several student teacher networks that are ordered as T1 to S2 and S1 to T2 through an embedding that serves the role as a softened interface, The first teacher student network of T1 to S2 is cascaded in sequence with an embedding. The second student teacher network is ordered as S1 to T2 with an embedding in a cascaded manner. The system cross knowledge distillation framework allows for compressing of the teacher network to create an efficient student network from a pre-trained teacher network, which is taught in Section 3 title Proposed Method on pages 163-165.);
(2) acquiring a low-definition image of a training set, wherein the training set is an open data set and comprises low-definition images and high-definition images in pairs; inputting the low-definition image into the two optimized paths for feature extraction to obtain a texture detail, a high-frequency feature and an image structure of the low-definition image; generating a preliminary image result from an extracted feature map through a convolution network, and calculating a loss term based on the preliminary image result and the high-definition image to punish an incorrect image restoration by the network, wherein the loss term comprises whether the image structure is consistent, whether color conforms to a statistical law and whether the texture detail is natural; and finally, optimizing parameters of the two optimized paths based on the loss term (e.g. as seen in figure 1, the system acquires a low definition image from a training set including high and low definition images. The system in figure 1 includes inputting the low definition image into two optimized paths for extraction of features. High frequency information is extracted similar to related work mentioned in section 2.2 under Knowledge Distillation on page 163. An image result is generated from the extracted feature maps detailed on page 164 through a convolution layer. Centered Kernal alignment is used to obtain similarity matrices to calculate the loss, which includes the details mentioned in Section 301 on page 163 and textures to gauge whether the structure of the images is consistent between networks mentioned on page 165 in the Result section. The loss related to reconstruction relates to the color conforming to a standard and the image appearing as close to the natural image as possible. Both of these losses are within a total loss calculated and detailed in equations (11) and (12) on page 165.); and
(3) taking out and connecting two cross-distilled student modules to form a final optimized path; inputting the low-definition image into the student network corresponding to the final optimized path; generating a super-resolution image for supervision using a complete teacher network, to calculate the loss term and update the parameters; obtaining a final compression model after training is complete and inputting the acquired low-definition image into the final compression model to obtain a super-resolution image (e.g. figure 1(b) discloses two cross distilled student modules connected to form a path in order to input a low definition image into the student network. The connected student modules are under the supervision of the original teacher’s super resolution (i.e. SR) images in order to generate a SR image. Loss is calculated in order to update the parameters within the connected modules. The eventual goal is to create a compact and efficient student network in order to input a low-definition image that is used to obtain a super-resolution image, which is mentioned in the proposed method section 3 on page 163..).
Re claim 2: (ORIGINAL) The image super-resolution method based on the knowledge distillation compression model according to claim 1, wherein a purpose of knowledge transfer is achieved by aligning respective inter-layer relationship matrices of teachers and students; and a specific processing flow of output features X and Y of different layers is as follows:
PNG
media_image1.png
132
262
media_image1.png
Greyscale
where tr(*) denotes a trace of a matrix, T denotes a transposition of a matrix; AT and As denote an inter-layer relationship matrix of the teacher network and the student network, respectively; and ||*||1 denotes L1 regularization (e.g. as seen on page 164, the equations (4)-(7) represent the description of aligning the inter-layer relationship matrices of the teachers and students using features of X and Y. This is explained in section 3.2 CKA Based Similarity Loss.).
Re claim 3: (CURRENTLY AMENDED) The image super-resolution method based on the knowledge distillation compression model according to claim 1, wherein a softened interface is added at a cross cascade, a feature matrix is linearly mapped by using k learnable parameters to complete dimension matching between teachers and students; a softened interface is designed by using channel separation and residual connection, the softened interface preserves a low- frequency image contour through residual to prevent an image edge from blurring and a gradient of the network from disappearing, the softened interface is used to further extract a high-frequency texture through a 3*3 convolution layer, and is used for a smooth transfer of knowledge between the teacher network and the student network; and the softened interface integrates the semantic features of a previous layer with the features extracted from the a current layer as a transition between teachers and students (e.g. the embedding is used as the softened interface that is added at a cross cascade. Feature matrices are formed and linearly mapped using learnable parameters in order to utilize for loss calculation. The embedding ensures there is no mismatch between modules using both residual connection and being on separate channels. This is seen in figure 1(a) and discussed on page 164 in the first paragraph. With ensuring there is no mismatch, it preserves the low-frequency image, including its contour. The embedding includes a layer that can extract features from the image and ensure the input of the teacher/student is mapped to the output of the prior student/teacher module. In addition, the prior layer information is integrated with the current layer module with the use of the embedding module, which is seen in figure 1 and explained in sections 3.1 The Cross Distillation Framework and 3.2 CKA Based similarity Loss sections on pages 163 and 164.).
Re claim 4: (ORIGINAL) The image super-resolution method based on the knowledge distillation compression model according to claim 1, wherein in a cross distillation stage, the student network is optimized by minimizing reconstruction loss and similarity loss; and in an integration distillation stage, the student network is optimized by minimizing the reconstruction loss with teachers (e.g. as seen in figure 1(a), the system minimizes the loss associated with the reconstruction and similarity and figure 1(b) shows the minimization of the reconstruction loss with the teacher supervision, which is explained in sections 3.1 The Cross Distillation Framework and 3.2 CKA Based similarity Loss sections on pages 163 and 164.).
Re claim 7: (PREVIOUSLY PRESENTED) Fang et al discloses the image super-resolution device based on the knowledge distillation compression model according to claim 5, wherein a purpose of knowledge transfer is achieved by aligning respective inter-layer relationship matrices of teachers and students; and a specific processing flow of output features X and Y of different layers is as follows:
PNG
media_image1.png
132
262
media_image1.png
Greyscale
where tr(*) denotes a trace of a matrix, T denotes a transposition of a matrix; AT and As denote an inter-layer relationship matrix of the teacher network and the student network, respectively; and |*I|1 denotes L1 regularization (e.g. as seen on page 164, the equations (4)-(7) represent the description of aligning the inter-layer relationship matrices of the teachers and students using features of X and Y. This is explained in section 3.2 CKA Based Similarity Loss.).
Re claim 8: (CURRENTLY AMENDED) Fang et al discloses the image super-resolution device based on the knowledge distillation compression model according to claim 5, wherein a softened interface is added at a cross cascade, a feature matrix is linearly mapped by using k learnable parameters to complete dimension matching between teachers and students; a softened interface is designed by using channel separation and residual connection, the softened interface preserves a low- frequency image contour through residual to prevent an image edge from blurring and a gradient of the network from disappearing, the softened interface is used to further extract a high-frequency texture through a 3*3 convolution layer, and is used for a smooth transfer of knowledge between the teacher network and the student network; and the softened interface integrates the semantic features of the a previous layer with the features extracted from the a current layer as a transition between teachers and students (e.g. the embedding is used as the softened interface that is added at a cross cascade. Feature matrices are formed and linearly mapped using learnable parameters in order to utilize for loss calculation. The embedding ensures there is no mismatch between modules using both residual connection and being on separate channels. This is seen in figure 1(a) and discussed on page 164 in the first paragraph. With ensuring there is no mismatch, it preserves the low-frequency image, including its contour. The embedding includes a layer that can extract features from the image and ensure the input of the teacher/student is mapped to the output of the prior student/teacher module. In addition, the prior layer information is integrated with the current layer module with the use of the embedding module, which is seen in figure 1 and explained in sections 3.1 The Cross Distillation Framework and 3.2 CKA Based similarity Loss sections on pages 163 and 164.).
Re claim 9: (PREVIOUSLY PRESENTED) Fang et al discloses the image super-resolution device based on the knowledge distillation compression model according to claim 5, wherein in a cross distillation stage, the student network is optimized by minimizing reconstruction loss and similarity loss; and in an integration distillation stage, the student network is optimized by minimizing the reconstruction loss with teachers (e.g. as seen in figure 1(a), the system minimizes the loss associated with the reconstruction and similarity and figure 1(b) shows the minimization of the reconstruction loss with the teacher supervision, which is explained in sections 3.1 The Cross Distillation Framework and 3.2 CKA Based similarity Loss sections on pages 163 and 164.).
Re claim 11: (CURRENTLY AMENDED) Fang et al discloses the non-transitory computer-readable storage medium in which one or more computer programs are stored according to claim 10, wherein a purpose of knowledge transfer is achieved by aligning respective inter-layer relationship matrices of teachers and students; and a specific processing flow of output features X and Y of different layers is as follows
PNG
media_image1.png
132
262
media_image1.png
Greyscale
where tr(*) denotes a trace of a matrix, T denotes a transposition of a matrix; AT and As denote an inter-layer relationship matrix of the teacher network and the student network, respectively; and |*I|1 denotes L1 regularization (e.g. as seen on page 164, the equations (4)-(7) represent the description of aligning the inter-layer relationship matrices of the teachers and students using features of X and Y. This is explained in section 3.2 CKA Based Similarity Loss.).
Re claim 12: (CURRENTLY AMENDED) Fang et al discloses the non-transitory computer-readable storage medium in which one or more computer programs are stored according to claim 10, wherein a softened interface is added at a cross cascade, a feature matrix is linearly mapped by using k learnable parameters to complete dimension matching between teachers and students; a softened interface is designed by using channel separation and residual connection, the softened interface preserves a low- frequency image contour through residual to prevent an image edge from blurring and a gradient of the network from disappearing, the softened interface is used to further extract a high-frequency texture through a 3*3 convolution layer, and is used for a smooth transfer of knowledge between the teacher network and the student network; and the softened interface integrates the semantic features of the a previous layer with the features extracted from the a current layer as a transition between teachers and students (e.g. the embedding is used as the softened interface that is added at a cross cascade. Feature matrices are formed and linearly mapped using learnable parameters in order to utilize for loss calculation. The embedding ensures there is no mismatch between modules using both residual connection and being on separate channels. This is seen in figure 1(a) and discussed on page 164 in the first paragraph. With ensuring there is no mismatch, it preserves the low-frequency image, including its contour. The embedding includes a layer that can extract features from the image and ensure the input of the teacher/student is mapped to the output of the prior student/teacher module. In addition, the prior layer information is integrated with the current layer module with the use of the embedding module, which is seen in figure 1 and explained in sections 3.1 The Cross Distillation Framework and 3.2 CKA Based similarity Loss sections on pages 163 and 164.).
Re claim 13: (CURRENTLY AMENDED) Fang et al discloses the non-transitory computer-readable storage medium in which one or more computer programs are stored according to claim 10, wherein in a cross distillation stage, the student network is optimized by minimizing reconstruction loss and similarity loss; and in an integration distillation stage, the student network is optimized by minimizing the reconstruction loss with teachers (e.g. as seen in figure 1(a), the system minimizes the loss associated with the reconstruction and similarity and figure 1(b) shows the minimization of the reconstruction loss with the teacher supervision, which is explained in sections 3.1 The Cross Distillation Framework and 3.2 CKA Based similarity Loss sections on pages 163 and 164.).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 5 and 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Li (US Pub 2022/0004803).
Re claim 5: (PREVIOUSLY PRESENTED) However, Fang et al fails to specifically teach the features of an image super-resolution device based on a knowledge distillation compression model, comprising: one or more processors; a memory for storing one or more programs; when executed by the one or more processors, the one or more programs cause the one or more processors to execute the image super-resolution method based on the knowledge distillation compression model according to claim 1.
However, this is well known in the art as evidenced by Li. Similar to the primary reference, Li discloses student teacher networks (same field of endeavor or reasonably pertinent to the problem).
Li discloses one or more processors; a memory for storing one or more programs; when executed by the one or more processors, the one or more programs cause the one or more processors to execute the image super-resolution method based on the knowledge distillation compression model according to claim 1 (e.g. the system discloses a computing device that contains a storage unit that executes a program to perform the invention. Incorporating a computing device with a storage unit to perform the features of the primary reference performs the features of the claims.).
[0023] In more detail, GANs have presented significant potential in modeling high dimensional distributions of image data, on a variety of visual tasks. Many of these tasks, such as style-transfer [31, 15] and super-resolution [17], are considered to be image-to-image translation tasks, in which we train a model to map images from one domain to another. (It will be understood that a meaningful feature in an image defines an “attribute”, such as, hair colour, age, gender, facial expression, etc. and that an “attribute value” is a particular example of a value of an attribute (e.g. blond/black/brown/red, 20/50, male/female, happy/sad/angry, etc.). A “domain” then is a set of images sharing a same attribute value.) The community has shown success in researching solutions to generate high fidelity images [1, 26] and dealing with unpaired data [31]. The success in these works has also led to a popular trend of developing mobile applications based on generative models. However, little work has been done in making these models efficient on mobile devices. As a result, the state-of-the-art GAN models are often large and slow on resource-limited edge devices. For instance, a CycleGAN [31] model needs 2.69 seconds to process one image of resolution 256×256 on a single CPU core of Intel® Xeon® E5-2686, with the model being 44M large.
[0093] Thus, any of the computing devices, but particularly the mobile devices, provide a computing device to translate an image from a first domain space to a second domain space. The computing device comprises a storage unit storing a generative adversarial network (GAN)-based student generator (G.sub.S), configured to translate images from the first domain space to the second domain space, the generator G.sub.S scaled downwardly in relation to a GANs-based teacher generator (G.sub.T) also configured to translate images from the first domain space to the second domain space. In an embodiment, the generator G.sub.S comprises a configuration (e.g. that is conditioned) to maintain a semantic relation knowledge determined for the generator G.sub.T. In an embodiment, the computing device comprises a processing unit configured to (e.g. via the AR application 1120): receive the image; provide the image to the generator G.sub.S to obtain a translated image; and provide the translated image for presenting.
[0101] The network model training environment 1102 provides a computing device configured to perform a method such as a method to configure by conditioning a GANs-based student generator. It will be understood that embodiments of the computing device aspect of the network model training environment 1102 and any related embodiments of, for example, the student generator or model, apply to the training method aspect, with suitable adaptation. FIG. 12 showing a flow chart of operations 1200 in accordance with an embedment herein. The operations provide a method such as for training (e.g through condition) in an embodiment. At step 1202, operations configure through conditioning a generative adversarial network (GAN)-based student generator (G.sub.S) to translate images from a first domain space to a second domain space, the generator G.sub.S scaled downwardly in relation to a GANs-based teacher generator (G.sub.T) configured to translate images from the first domain space to the second domain space, wherein the generator G.sub.S is conditioned using knowledge distillation to maintain a semantic relation knowledge determined for the generator G.sub.T. At step 1204, operations provide the generator G.sub.S for use by a computing device to translate an image. A related computed device and computer program product aspect will be apparent as will other aspects.
Therefore, in view of Li, it would have been obvious to one of ordinary skill before the effective filing date of the claimed invention was made to have the feature of one or more processors; a memory for storing one or more programs; when executed by the one or more processors, the one or more programs cause the one or more processors to execute the image super-resolution method based on the knowledge distillation compression model according to claim 1, incorporated in the device of Fang et al, in order to utilize knowledge distillation on computing devices, which reduces parameters, footprint and inference time execution on devices (as stated in Li ¶ [80]).
Re claim 10: (CURRENTLY AMENDED) However, Fang fails to specifically teach the features of a non-transitory computer-readable storage medium in which one or more computer programs are stored, wherein the one or more computer programs comprise program codes which are used to execute the image super-resolution method based on the knowledge distillation compression model according to claim 1 when the computer programs are run on a computer.
However, this is well known in the art as evidenced by Li. Similar to the primary reference, Li discloses student teacher networks (same field of endeavor or reasonably pertinent to the problem).
Li discloses program codes which are used to execute the image super-resolution method based on the knowledge distillation compression model according to claim 1 when the computer programs are run on a computer (e.g. the system discloses a computing device that contains a storage unit that executes a program to perform the invention. Incorporating a computing device with a storage unit to perform the features of the primary reference performs the features of the claims.).
Therefore, in view of Li, it would have been obvious to one of ordinary skill before the effective filing date of the claimed invention was made to have the feature of program codes which are used to execute the image super-resolution method based on the knowledge distillation compression model according to claim 1 when the computer programs are run on a computer, incorporated in the device of Fang et al, in order to utilize knowledge distillation on computing devices, which reduces parameters, footprint and inference time execution on devices (as stated in Li ¶ [80]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Kim discloses student-teacher transfer learning.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHAD S DICKERSON whose telephone number is (571)270-1351. The examiner can normally be reached Monday-Friday 10AM-6PM 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, Abderrahim Merouan can be reached at 571-270-5254. 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.
/CHAD DICKERSON/ Primary Examiner, Art Unit 2683
Prosecution Insights