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 Amendment
Claims 1-24 were pending for examination in the Application no. 18/855,943 filed October 10th, 2024. In the preliminary remarks and amendments received on May 19th, 2025, claims 19, 21, 23, and 24 are cancelled. Accordingly, claims 1-18, 20, and 22 are currently pending for examination in the application.
Priority
Acknowledgment is made of applicant’s status as a U.S. National Stage Filing under 35 U.S.C. § 371 of International Application No. , filed on May 16th, 2022.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on October 10th, 2024, is in compliance with the provisions of 37 CFR 1.97. Accordingly, the IDS is being considered and attached by the examiner.
Claim Objections
Claim 4 is objected to because of the following informalities:
In claim 4, “the plurality of size-specific segmentation prediction” should be “the plurality of size-specific segmentation [[prediction]]predictions”;
Appropriate correction is required.
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.
Claims 1-8 are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (Wang-1; “Resolution Switchable Networks for Runtime Efficient Image Recognition,” 2020) in view of Wang et al. (Wang-2; “Supplementary Material for ‘Resolution Switchable Networks for Runtime Efficient Image Recognition’,” 2020).
Regarding claim 1, Wang-1 discloses a method comprising:
resizing a training image to a plurality of training images at different image resolutions (section 3.1 on pg. 536 and 2nd para. of section 3.1 on pg. 537, recite(s)
[section 3.1 on pg. 536] “To make the description self-contained, we begin with the basic training of a CNN model. Given training samples, we crop and resize each sample to a fixed resolution image
x
i
. …”
[2nd para. of section 3.1 on pg. 537] “In this part, we propose multi-resolution parallel training, or called parallel training for brevity, to train a single model which can switch image resolutions at runtime. During training, each image sample is randomly cropped and resized to several duplicate images with different resolutions. …”
, where “each image sample” is a training image “resized” at “different [image] resolutions”);
generating a plurality of size-specific(Fig. 2 recite(s):
PNG
media_image1.png
400
819
media_image1.png
Greyscale
, where the “corresponding prediction” for each corresponding image of “different resolutions” are size-specific outputs generated by applying each of the training images (e.g., “Images of different resolutions are trained…”) to a computer model (e.g., “RS-Net”) to generate an associated size-specific prediction (e.g., “corresponding prediction
p
s
”) for each training image resolution (e.g., “according to the image resolution”)), the computer model having a shared convolutional layer with the same parameters applied to each image resolution (2nd para. of section 3.1 on pg. 537 further recite(s):
[2nd para. of section 3.1 on pg. 537] “ …In CNNs, if input images have different resolutions, the corresponding feature maps in all Conv layers will also vary in resolution. Thanks to GAP, features are transformed to a unified spatial dimension (1×1) with equal amount of channels, making it possible to be followed by a same FC layer. During our parallel training, we share parameters of Conv layers and the FC layer, and therefore the training for multiple resolutions can be realized in a single network. …”
, where the “share[d] parameters of Conv layers” are shared convolutional layers with the same parameters applied to each image resolution (i.e., “multiple resolutions”)); and
training parameters of the computer model, including parameters of the shared convolutional layer, based on a comparison of the plurality of size-specific(Fig. 2—see citation above—, where the 1st para. of section 3.1 on pgs. 536-537, 2nd para. of pg. 537, and 2nd para. of pg. 541 further recite(s):
PNG
media_image2.png
376
849
media_image2.png
Greyscale
PNG
media_image3.png
198
814
media_image3.png
Greyscale
PNG
media_image4.png
180
813
media_image4.png
Greyscale
, where optimizing the computer model “with a cross-entropy loss” is training (e.g., “optimiz[ing]” and/or “updat[ing]”) parameters of the computer model (e.g., “network configurations with parameters
θ
” and/or “network weights”), including parameters of the shared convolutional layer (e.g., “parameters of Conv layers and the FC layer…”), based on a comparison of the plurality of size-specific predictions (e.g., “predicted probability of the class
c
… denoted as
p
(
c
|
x
i
,
θ
)
” or “
p
s
” in Fig. 2) with a label of the training image (e.g., the “ground truth[s]” “
y
i
” or the “labels” depicted in Fig. 2)).
Where Wang-1 does not specifically disclose
generating a plurality of size-specific segmentation outputs by applying each of the training images to a computer model to generate an associated size-specific segmentation prediction for each training image resolution…; and
training parameters of the computer model… based on a comparison of the plurality of size-specific segmentation predictions with a label of the training image;
Wang-2 teaches in the same field of endeavor generating a plurality of size-specific outputs by applying each of training images to a computer model
generating a plurality of size-specific segmentation outputs by applying each of the training images to a computer model to generate an associated size-specific segmentation prediction for each training image resolution (Fig. 2 and section 3 on pg. 1, recite(s)
PNG
media_image5.png
482
1134
media_image5.png
Greyscale
[section 3 on pg. 1] “Besides the experiments described in the main paper, we also apply our method to semantic segmentation to further validate its generalization ability to handle other visual recognition tasks beyond classification. …”
, where the predictions
p
s
,
s
∈
{
1,2
,
…
,
S
}
depicted in Fig. 2 are a plurality of size-specific segmentation outputs by applying each of training images (e.g.,
x
1
,
x
2
,
…
,
x
S
depicted in Fig. 2) to a computer model (e.g., RS-Net) to generate an associated size-specific segmentation predictions (e.g.,
p
s
,
s
∈
{
1,2
,
…
,
S
}
) for each training image resolution)…, and
training parameters of the computer model… based on a comparison of the plurality of size-specific segmentation predictions with a label of the training image (Fig. 2—see citation above—, where the 1st para. on pg. 2 further recite(s):
PNG
media_image6.png
344
1129
media_image6.png
Greyscale
, where the “calculating losses with labels” is training parameters of the computer model (i.e., loss functions train parameters of the computer model as disclosed previously in Wang-1 above— see similar claim limitation above) based on at least a comparison of the plurality of size-specific segmentation predictions (e.g., predictions
p
s
,
s
∈
{
1,2
,
…
,
S
}
depicted in Fig. 2) with a label of the training image (e.g., “labels”)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the presently filed invention to modify the system of Wang-1 to apply their computer model to the task of semantic segmentation to generate a plurality of size-specific segmentation outputs and associated size-specific segmentation prediction for each training image resolution such that the training model of Wang-1 generates a plurality of size-specific segmentation outputs by applying each of the training images to a computer model to generate an associated size-specific segmentation prediction for each training image resolution and training parameters of the computer model, including parameters of the shared convolutional layer, based on a comparison of the plurality of size-specific segmentation predictions with a label of the training image because Wang-2 teaches that the computer model of Wang-1 can be applied to the task of semantic segmentation as detailed above.
Regarding claim 2, Wang-1 in view of Wang-2 discloses the method of claim 1, wherein Wang-1 further discloses the computer model includes one or more size-dependent layers with size-specific parameters (1st para. of pg. 535, recite(s)
[1st para. of pg. 535] “First, we propose a parallel training framework where images with different resolutions are trained within a single model. As the resolution difference usually leads to the difference of activation statistics in a network [29], we adopt shared network parameters but privatized Batch Normalization layers (BNs) [12] for each resolution. Switching BNs enables the model to flexibly switch image resolutions, without needing to adjust other network parameters.”
, where the “Batch Normalization layers (BNs)” are size-dependent layers with size-specific parameters).
Regarding claim 3, Wang-1 in view of Wang-2 discloses the method of claim 2, wherein Wang-1 further discloses the one or more size-dependent layers with size-specific parameters are normalization layers (1st para. of pg. 535—see citation in claim 2 above—, where the “Batch Normalization layers (BNs)” are normalization layers).
Regarding claim 4, Wang-1 in view of Wang-2 discloses the method of claim 1, wherein Wang-1 further discloses training the parameters of the computer model includes determining an ensemble segmentation prediction based on the plurality of size-specific segmentation predictions and training the parameters is based on reducing a training loss that includes a loss of the ensemble segmentation prediction (3rd-4th paras. of pg. 540, recite(s)
PNG
media_image7.png
417
818
media_image7.png
Greyscale
, where the “ensemble loss” is a loss of an ensemble prediction; wherein the ensemble prediction is based on the plurality of size-specific predictions (e.g., “ensemble prediction[s]”)).
Regarding claim 5, Wang-1 in view of Wang-2 discloses the method of claim 4, wherein Wang-1 further discloses training the parameters of the computer model includes training weights of the respective plurality of size-specific segmentation predictions for determining the ensemble segmentation prediction (3rd-4th paras. of pg. 540—see citation in claim 4 above—, where the “scores
α
=
[
α
1
α
2
…
α
S
]
” are training weights (hence the term “weighted”) of the respective plurality of size-specific segmentation predictions).
Regarding claim 6, Wang-1 in view of Wang-2 discloses the method of claim 1, wherein Wang-1 further discloses training the parameters of the computer model includes a distillation loss of a sequence of teaching labels based on an order of the respective image resolutions of the plurality of size-specific segmentation predictions (last para. of pg. 540 to first para. of pg. 541, and 3rd para. of pg. 541, recite(s)
PNG
media_image8.png
326
851
media_image8.png
Greyscale
PNG
media_image9.png
368
814
media_image9.png
Greyscale
, where the “distillation loss” is a distillation loss of a sequence of teaching labels (e.g., “teacher prediction
p
t
” for each “index
t
”) based on an order of the respective image resolutions of the plurality of size-specific segmentation predictions (e.g., “guidance from predications at high resolutions toward predictions at lower resolutions”)).
Regarding claim 7, Wang-1 in view of Wang-2 discloses the method of claim 6, wherein Wang-1 further discloses the distillation loss includes an ensemble segmentation prediction as a first teacher in the sequence of teaching labels (last para. of pg. 540 to first para. of pg. 541—see citation in claim 6 above—, where the “teacher prediction
p
t
” is an ensemble prediction as a first teacher in the sequence of teaching labels).
Regarding claim 8, Wang-1 in view of Wang-2 discloses the method of claim 1, wherein Wang-2 further teaches the size-specific segmentation predictions are resized to a maximum image resolution for comparison to the image label (1st para. on pg. 2—see citation in the teaching of claim 1 limitation “training parameters…” above—, where “resiz[ing]
z
^
s
to the largest input resolution… before ensemble distillation process and calculating losses with labels” is resizing the size-specific segmentation predictions to a maximum image resolution for comparison to the image label (i.e. since the size-specific segmentation predictions
p
s
,
s
∈
{
1,2
,
…
,
S
}
are based on the logits
z
^
s
, which are resized to the maximum image resolution, the size-specific segmentation predictions will also be resized to a maximum image resolution)).
Claims 9-18, 20, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Wang-1 in view of Wang-2 as applied to claims 1-8 above, and further in view of Wang et al. (Wang-3; US 2019/0164290 A1, provided by Applicant in IDS filed October 10th, 2024).
Regarding claim 9, the claim recites similar limitations to claim 1 but in the form of a system comprising: a processor; and a non-transitory computer-readable storage medium containing computer program for execution by the processor for performing the method of claim 1. Therefore, claim 9 recites similar limitations to claim 1 and is rejected for similar rationale and reasoning (see the analysis for claim 1 above).
Where Wang-1 in view of Wang-2 does not specifically disclose
a system comprising: a processor; and a non-transitory computer-readable storage medium containing computer program for execution by the processor…;
Wang-3 discloses in the same field of endeavor of image segmentation
a system comprising: a processor; and a non-transitory computer-readable storage medium containing computer program for execution by the processor… (the abstract and para(s). [0016], recite(s)
[abstract] “Techniques related to implementing fully convolutional networks for semantic image segmentation are discussed. …”
[0016] “The material disclosed herein may be implemented in hardware, firmware, software, or any combination thereof. The material disclosed herein may also be implemented as instructions stored on a machine-readable medium, which may be read and executed by one or more processors. A machine-readable medium may include any medium and/or mechanism for storing or transmitting information in a form readable by a machine (e.g., a computing device). For example, a machine-readable medium may include read only memory (ROM); random access memory (RAM); magnetic disk storage media; optical storage media; flash memory devices…”
, where the machine-readable medium comprises of at least non-transitory computer-readable storage medium (e.g., “ROM”, “RAM”, etc.)).
Since Wang-2 and Wang-3 each disclose applying a computer model (e.g., convolutional neural networks) to the task of semantic segmentation, it would have been obvious to one of ordinary skill in the art before the effective filing date of the presently filed invention to modify the system of Wang-1 in view of Wang-2 to incorporate a processor and a non-transitory computer-readable storage medium containing computer program to implement the computer method of Wang-1 in view of Wang-2 as taught by Wang-3 above.
Regarding claim 10, the claim recites similar limitations to claim 2 and is rejected for similar rationale and reasoning (see the analysis for claim 2 above).
Regarding claim 11, the claim recites similar limitations to claim 3 and is rejected for similar rationale and reasoning (see the analysis for claim 3 above).
Regarding claim 12, the claim recites similar limitations to claim 4 and is rejected for similar rationale and reasoning (see the analysis for claim 4 above).
Regarding claim 13, the claim recites similar limitations to claim 5 and is rejected for similar rationale and reasoning (see the analysis for claim 5 above).
Regarding claim 14, the claim recites similar limitations to claim 6 and is rejected for similar rationale and reasoning (see the analysis for claim 6 above).
Regarding claim 15, the claim recites similar limitations to claim 7 and is rejected for similar rationale and reasoning (see the analysis for claim 7 above).
Regarding claim 16, the claim recites similar limitations to claim 8 and is rejected for similar rationale and reasoning (see the analysis for claim 8 above).
Regarding claim 17, the claim recites the non-transitory computer-readable storage medium of claim 9 above. Therefore, claim 17 recites similar limitations to claim 9 and is rejected for similar rationale and reasoning (see the analysis for claim 9 above).
Regarding claim 18, the claim recites similar limitations to claim 10 and is rejected for similar rationale and reasoning (see the analysis for claim 10 above).
Regarding claim 20, the claim recites similar limitations to claim 12 and is rejected for similar rationale and reasoning (see the analysis for claim 12 above).
Regarding claim 22, the claim recites similar limitations to claim 14 and is rejected for similar rationale and reasoning (see the analysis for claim 14 above).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JULIA Z YAO whose telephone number is (571)272-2870. The examiner can normally be reached Monday - Friday (8:30AM - 5PM).
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, Emily Terrell can be reached at (571)270-3717. 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.
/J.Z.Y./Examiner, Art Unit 2666
/MING Y HON/Primary Examiner, Art Unit 2666
Prosecution Insights