DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 12/19/2025 has been entered. Claims 4-5, 7, 11 and 14-15 have been canceled, claim 17-26 have been added. Claims 6, 8-10, 12-13, 16-17 and 25 have been allowed, claims 1-3, 18-24 and 26 remain pending in the application.
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 of this title, 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.
Claim 1-3, 18-20 and 22-24 are rejected under 35 U.S.C. 103 as being unpatentable over Gandelsman U.S. Patent Application 20240169604 in view of Xia U.S. Patent Application 20210342977.
Regarding claim 1, Gandelsman discloses a computing system configured to perform image generation, the computing system comprising:
one or more processors (processor(s) 1405); and
one or more non-transitory computer-readable media (memory 1410) that collectively store:
a machine-learned denoising diffusion model, wherein the machine-learned denoising diffusion model (paragraph [0105]: at each denoising step, from the intermediate output image of the diffusion model, machine learning model 225 extracts each entity region using the selection region) comprises:
a first downsampling block configured to process an input and perform a first downsampling operation to generate a first downsampled output; a second downsampling block configured to process the first downsampled output and perform a second downsampling operation to generate a second downsampled output (paragraph [0076]: The U-Net 500 takes input features 505 having an initial resolution and an initial number of channels, and processes the input features 505 using an initial neural network layer 510 (e.g., a convolutional network layer) to produce intermediate features 515. The intermediate features 515 are then down-sampled using a down-sampling layer 520 such that down-sampled features 525 features have a resolution less than the initial resolution and a number of channels greater than the initial number of channels; paragraph [0077]: This process is repeated multiple times; see fig. 5);
a first upsampling block configured to process the output and perform a first upsampling operation to generate a first upsampling output; and a second upsampling block configured to process the first upsampling output and perform a second upsampling operation to generate a second upsampling output (paragraph [0077]: the process is reversed. For example, the down-sampled features 525 are up-sampled using up-sampling process 530 to obtain up-sampled features 535. The up-sampled features 535 can be combined with intermediate features 515 having a same resolution and number of channels via a skip connection 540. These inputs are processed using a final neural network layer 545 to produce output features 550; see fig. 5); and
instructions that, when executed by the one or more processors, cause the computing system to use the machine-learned denoising diffusion model configured to generate images (paragraph [0088]: At operation 615, the system generates an output image based on the encoding; paragraph [0089]: At operation 620, the system displays the output image to the user; paragraph [0105]: at each denoising step, from the intermediate output image of the diffusion model, machine learning model 225 extracts each entity region using the selection region).
Gandelsman discloses all the features with respect to claim 1 as outlined above. However, Gandelsman fails to disclose a transformer block configured to perform self-attention on the output to generate a transformer output, wherein the transformer block performs self-attention on the second downsampled output only after the first downsampling operation.
Xia discloses a transformer block configured to perform self-attention on the output to generate a transformer output, wherein the transformer block performs self-attention on the second downsampled output only after the first downsampling operation (paragraph [0103]: the input end of the self-attention module (transformer block) can be connected with an output end of a second residual module (second downsampled output), and the output end of the self-attention module (transformer output) can be connected with an input end of a third residual module. In other words, the self-attention module is disposed behind the second down-sampling residual module).
Therefore, it would have been obvious before the effective filing date of the claimed invention to combine Gandelsman’s to use a transformer model as taught by Xia, to reduce computational complexity and extract global features well.
Regarding claim 2, Gandelsman as modified by Xia discloses the computing system of claim 1, wherein the first downsampling block and the second upsampling block do not perform self-attention (Gandelsman’s paragraph [0076]: The U-Net 500 takes input features 505 having an initial resolution and an initial number of channels, and processes the input features 505 using an initial neural network layer 510 (e.g., a convolutional network layer) to produce intermediate features 515. The intermediate features 515 are then down-sampled using a down-sampling layer 520 such that down-sampled features 525 features have a resolution less than the initial resolution and a number of channels greater than the initial number of channels; paragraph [0077]: This process is repeated multiple times, and then the process is reversed; see fig. 5; Xia’s paragraph [0103]: the self-attention module is disposed behind the second down-sampling residual module).
Therefore, it would have been obvious before the effective filing date of the claimed invention to combine Gandelsman’s to use a transformer model as taught by Xia, to reduce computational complexity and extract global features well.
Regarding claim 3, Gandelsman as modified by Xia discloses the computing system of claim 1, wherein the first downsampling block and the second upsampling block perform only convolutional operations (Gandelsman’s paragraph [0076]: processes the input features 505 using an initial neural network layer 510 (e.g., a convolutional network layer) to produce intermediate features 515. The intermediate features 515 are then down-sampled using a down-sampling layer 520 such that down-sampled features 525 features have a resolution less than the initial resolution and a number of channels greater than the initial number of channels; paragraph [0077]: This process is repeated multiple times, and then the process is reversed).
Therefore, it would have been obvious before the effective filing date of the claimed invention to combine Gandelsman’s to use a transformer model as taught by Xia, to reduce computational complexity and extract global features well.
Claim 18 recites the functions of the apparatus recited in claim 1 as method steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 1 applies to the method steps of claim 18.
Claim 19 recites the functions of the apparatus recited in claim 2 as method steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 2 applies to the method steps of claim 19.
Claim 20 recites the functions of the apparatus recited in claim 3 as method steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 3 applies to the method steps of claim 20.
Claim 22 recites the functions of the apparatus recited in claim 1 as medium steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 1 applies to the medium steps of claim 22.
Claim 23 recites the functions of the apparatus recited in claim 2 as medium steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 2 applies to the medium steps of claim 23.
Claim 24 recites the functions of the apparatus recited in claim 3 as medium steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 3 applies to the medium steps of claim 24.
Claim 21 and 26 are rejected under 35 U.S.C. 103 as being unpatentable over Gandelsman U.S. Patent Application 20240169604 in view of Xia U.S. Patent Application 20210342977, and further in view of Tran U.S. Patent 10984245.
Regarding claim 26, Gandelsman as modified by Xia all the features with respect to claim 1 as outlined above. However, Gandelsman as modified by Xia fails to disclose the machine-learned model sequentially applies channel multipliers of 1, 2, and 4.
Tran discloses the machine-learned model sequentially applies channel multipliers of 1, 2, and 4 (col. 17 line 32-35: The networks convolution layer channels are uniformly multiplied by different multipliers, e.g., {0.5, 0.75, 1, 1.25, 1.5, 1.75, 2, 4}, which results in networks with the same depth but different width).
Therefore, it would have been obvious before the effective filing date of the claimed invention to combine Gandelsman and Xia’s to use different channel size as taught by Tran, to improve video analysis using machine-learning algorithms within a social-networking environment.
Claim 21 recites the functions of the apparatus recited in claim 26 as method steps. Accordingly, the mapping of the prior art to the corresponding functions of the apparatus in claim 26 applies to the method steps of claim 21.
Allowable Subject Matter
Claims 6, 8-10, 12-13, 16-17 and 25 are allowed.
The following is an examiner’s statement of reasons for allowance:
Claim 6 is about obtaining data descriptive of a specified resolution for synthetic images to be generated by the machine-learned diffusion model; accessing a noise schedule generation algorithm that outputs an output noise schedule based on an input resolution, wherein an output signal to noise ratio associated with the output noise schedule is inversely correlated to a magnitude of the input resolution; performing the noise schedule generation algorithm with the specified resolution as the input resolution to generate a resolution-specific noise schedule for the machine-learned diffusion model; and employing the machine-learned diffusion model to generate the synthetic images of the specified resolution according to the resolution-specific noise schedule; wherein performing the noise schedule generation algorithm with the specified resolution as the input resolution comprises:
obtaining a reference signal to noise ratio associated with a reference noise schedule associated with a reference resolution; determining a scaling value based on the reference resolution and the specified resolution, wherein a magnitude of the scaling value is correlated to a ratio of the reference resolution to the specified resolution; and scaling the reference signal to noise ratio by the scaling value to obtain a resolution-specific signal to noise ratio for the resolution-specific noise schedule.
Gandelsman 20240169604, Chen 20240161327, and Yoon 20220221339 combined cannot discloses these limitations perfectly. These limitations when read in light of the rest of the limitations in the claim make the claim allowable subject matter.
Claim 8-10 and 12 depend on claim 6, are allowed based on same reason as claim 6.
Claim 13 and 25 are about improving training of diffusion models for image synthesis via multi-scale training loss comprising:
Accessing a training input associated with a training image, wherein the training input comprises a first epsilon value; employing a diffusion model to generate a predicted output associated with a synthetic image based at least in part on the training input associated with the training image, wherein the predicted output comprises a second epsilon value; evaluating a multi-scale loss function based on the training input associated with the training image and the predicted output associated with the synthetic image, wherein evaluating the multi-scale loss function comprises:
downsampling both the training input and the predicted output to a plurality of reduced resolutions; evaluating a plurality of loss values between the training input and the predicted output respectively at the plurality of reduced resolutions; and aggregating the plurality of loss values to generate an aggregate loss value, wherein aggregating the plurality of loss values to generate the aggregate loss value comprises: weighting each loss value by a weighting coefficient, wherein a magnitude of the weighting coefficient applied to the loss value at each reduced resolution is inversely correlated to the reduced resolution; and updating one or more parameters of the diffusion model based at least in part on the aggregate loss value.
Chen 20240161327, Chen 20190130275, Yoon 20220221339 and Cook 20170109656 combined cannot discloses these limitations perfectly. These limitations when read in light of the rest of the limitations in the claim make the claim allowable subject matter.
Claim 16 depend on claim 13, are allowed based on same reason as claim 13.
Claim 17 is about obtaining data descriptive of a specified resolution for synthetic images to be generated by the machine-learned diffusion model; accessing a noise schedule generation algorithm that outputs an output noise schedule based on an input resolution, wherein an output signal to noise ratio associated with the output noise schedule is inversely correlated to a magnitude of the input resolution; performing the noise schedule generation algorithm with the specified resolution as the input resolution to generate a resolution-specific noise schedule for the machine-learned diffusion model; and employing the machine-learned diffusion model to generate the synthetic images of the specified resolution according to the resolution-specific noise schedule; wherein performing the noise schedule generation algorithm with the specified resolution as the input resolution comprises:
obtaining a first reference signal to noise ratio associated with a first reference noise schedule associated with a first reference resolution; determining a first scaling value based on the first reference resolution and the specified resolution, wherein a magnitude of the first scaling value is correlated to a first ratio of the first reference resolution to the specified resolution; scaling the first reference signal to noise ratio by the first scaling value to obtain a first resolution-specific signal to noise ratio; obtaining a second reference signal to noise ratio associated with a second reference noise schedule associated with a second, different reference resolution; determining a second scaling value based on the second reference resolution and the specified resolution, wherein a magnitude of the second scaling value is correlated to a second ratio of the second reference resolution to the specified resolution; and scaling the second reference signal to noise ratio by the second scaling value to obtain a second resolution-specific signal to noise ratio; wherein the resolution-specific noise schedule comprises an interpolated schedule that interpolates between the first resolution-specific signal to noise ratio and the second resolution-specific signal to noise ratio.
Gandelsman 20240169604, Chen 20240161327, and Yoon 20220221339 combined cannot discloses these limitations perfectly. These limitations when read in light of the rest of the limitations in the claim make the claim allowable subject matter.
Response to Arguments
Applicant's arguments filed 12/19/2025, page 10, with respect to the rejection(s) of claim(s) 1 under 103, have been fully considered and are moot upon a new ground(s) of rejection made under 35 U.S.C. 103 as being unpatentable over Gandelsman U.S. Patent Application 20240169604 in view of Xia U.S. Patent Application 20210342977, as outlined above.
Applicant argues on page 10 that "In the interview, the Examiner indicated the amendments herein would place the application in condition for allowance”.
In reply, the applicant remove some limitation from claim 1, broaden the scope of claim, the examiner needs to do further search and consideration, and finds new art for a new ground of rejection. The applicant should add removed limitation back into independent claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Yi Yang whose telephone number is (571)272-9589. The examiner can normally be reached on Monday-Friday 9:00 AM-6:00 PM EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Daniel Hajnik can be reached on 571-272-7642. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/YI YANG/
Primary Examiner, Art Unit 2616