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 .
Continued Examination Under 37 CFR 1.114
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 04/03/2026 has been entered.
Response to Amendment
Applicant’s amendments filed on 04/03/2026 to the claims have overcome claim rejections under 35 U.S.C. 112(a) and under 35 U.S.C. 112(b) as preciously set forth in the Final Rejection Office Action mailed on 01/16/2026.
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.
Claim(s) 1-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Morvan et al (US 20230207106 A1), hereinafter Morvan in view of Laradji et al (Arxiv:1907.01430 2 July 2019), hereinafter Laradji.
-Regarding claim 1, Morvan discloses a method, comprising (Abstract; FIGS. 1-10
PNG
media_image1.png
477
235
media_image1.png
Greyscale
): updating (FIG. 6; [0053], “the NN may be implemented and trained in one of various ways”), using a bounding box in an input image as reference (FIG. 5, steps 500-502; [0051], “The selected point serves as a centroid of a bounding box having a size matching an input size of the NN. Preprocessing unit 200 may crop regions of the 3D image outside the box to generate a preprocessed 3D image”; [0095]; FIGS. 3-4), a first segmentation network to infer a first segmentation mask for an object in the input image (FIG. 5, step 504; [0059]; FIG. 2; [0048]; [0053]); and using the first segmentation mask as a conditional reference to infer a second segmentation mask for the input image (FIG. 5, steps 506-508; [0060]; FIG. 2; [0048]; [0054]).
Morvan does not disclose a second segmentation network, including a modified instance segmentation model that infers a second segmentation mask for the input image instead of generating a new bounding box, and a loss comparison with the second segmentation mask. However, Morvan does disclose refining segmentation mask by performing erosion and opening process on the voxel data generated from the first segmentation mask (FIG. 9).
In the same field of endeavor, Laradji teaches a method for instance segmentation by obtaining pseudo masks lead to a supervision for Mask R-CNN (Laradji: Abstract; FIGS. 1-5; Sec. 3.). Laradji teaches a second segmentation network, including a modified instance segmentation model that infers a second segmentation mask for the input image instead of generating a new bounding box, and a loss comparison with the second segmentation mask (Laradji : FIGS. 2-3
PNG
media_image2.png
512
858
media_image2.png
Greyscale
; Page 6, 3rd paragraph, “simply use the trained Mask R-CNN to predict the object masks
for an unseen image. To refine these masks, we leverage the same object proposal method
as that used in training.”).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Morvan with the teaching of Laradji by using a modified instance segmentation model to infer a second segmentation mask based on the loss comparison with the 1st segmentation mask in order to improve performance of segmentation without manually labeling (Laradji: Abstract; Sec. 1., last paragraph).
-Regarding claim 8, Morvan discloses a processor (FIG. 1, processing circuitry 103), comprising (Abstract; FIGS. 1-10): update (FIG. 6; [0053], “the NN may be implemented and trained in one of various ways”), using a bounding shape in an input image as reference (FIG. 5, steps 500-502; [0051], “The selected point serves as a centroid of a bounding box having a size matching an input size of the NN. Preprocessing unit 200 may crop regions of the 3D image outside the box to generate a preprocessed 3D image”; [0095]; FIGS. 3-4), a first segmentation network to infer a first segmentation for an object in the input image (FIG. 5, step 504; [0059]; FIG. 2; [0048]; [0053]); and using the first segmentation mask as a conditional reference to infer a second segmentation mask for the input image (FIG. 5, steps 506-508; [0060]; FIG. 2; [0048]; [0054]).
Morvan does not disclose a second segmentation network, including a modified instance segmentation model that infers a second segmentation mask for the input image instead of generating a new bounding shape, and a loss comparison with the second segmentation mask. However, Morvan does disclose refining segmentation mask by performing erosion and opening process on the voxel data generated from the first segmentation mask (FIG. 9).
In the same field of endeavor, Laradji teaches a method for instance segmentation by obtaining pseudo masks lead to a supervision for Mask R-CNN (Laradji: Abstract; FIGS. 1-5; Sec. 3.). Laradji teaches a second segmentation network, including a modified instance segmentation model that infers a second segmentation mask for the input image instead of generating a new bounding box, and a loss comparison with the second segmentation mask (Laradji : FIGS. 2-3; Page 6, 3rd paragraph, “simply use the trained Mask R-CNN to predict the object masks
for an unseen image. To refine these masks, we leverage the same object proposal method
as that used in training.”).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Morvan with the teaching of Laradji by using a modified instance segmentation model to infer a second segmentation mask based on the loss comparison with the 1st segmentation mask in order to improve performance of segmentation without manually labeling (Laradji: Abstract; Sec. 1., last paragraph).
-Regarding claim 15, Morvan discloses a system, comprising (Abstract; FIGS. 1-10): one or more processors (FIG. 1, processing circuitry 103) to update (FIG. 6; [0053], “the NN may be implemented and trained in one of various ways”), using a bounding box in an input image as reference (FIG. 5, steps 500-502; [0051], “The selected point serves as a centroid of a bounding box having a size matching an input size of the NN. Preprocessing unit 200 may crop regions of the 3D image outside the box to generate a preprocessed 3D image”; [0095]; FIGS. 3-4), a first segmentation network to infer a pseudo mask for an object in the input image (FIG. 5, step 504; [0059]; FIG. 2; [0048]; [0053]); and using the pseudo mask as a conditional reference to infer a second segmentation mask for the input image (FIG. 5, steps 506-508; [0060]; FIG. 2; [0048]; [0054]).
Morvan does not disclose a second segmentation network, including a modified instance segmentation model that infers a second segmentation mask for the input image instead of generating a new bounding box, and a loss comparison with the second segmentation mask. However, Morvan does disclose refining segmentation mask by performing erosion and opening process on the voxel data generated from the first segmentation mask (FIG. 9).
In the same field of endeavor, Laradji teaches a method for instance segmentation by obtaining pseudo masks lead to a supervision for Mask R-CNN (Laradji: Abstract; FIGS. 1-5; Sec. 3.). Laradji teaches a second segmentation network, including a modified instance segmentation model that infers a second segmentation mask for the input image instead of generating a new bounding box, and a loss comparison with the second segmentation mask (Laradji : FIGS. 2-3; Page 6, 3rd paragraph, “simply use the trained Mask R-CNN to predict the object masks
for an unseen image. To refine these masks, we leverage the same object proposal method
as that used in training.”).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Morvan with the teaching of Laradji by using a modified instance segmentation model to infer a second segmentation mask based on the loss comparison with the 1st segmentation mask in order to improve performance of segmentation without manually labeling (Laradji: Abstract; Sec. 1., last paragraph).
-Regarding claims 2 and 9, Morvan in view Laradji teaches the method of claim 1 and the processor of claim 8.
Morvan does not disclose pre-training the second segmentation network using a first set of segmentation mask.
In the same field of endeavor, Laradji teaches a method for instance segmentation by obtaining pseudo masks lead to a supervision for Mask R-CNN (Laradji: Abstract; FIGS. 1-5; Sec. 3.). Laradji further teaches pre-training the second segmentation network using a first set of segmentation mask (Laradji: FIGS. 2-3; Page 6, equation (2)).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Morvan with the teaching of Laradji by using a modified instance segmentation model to infer a second segmentation mask based on the loss comparison with the 1st segmentation mask in order to improve performance of segmentation without manually labeling (Laradji: Abstract; Sec. 1., last paragraph).
-Regarding claims 3, 10 and 16, , Morvan in view Laradji teaches the method of claim 1, the processor of claim 8, and the system of claim 15. The combination further teaches wherein the second segmentation network is a modified mask region-based convolutional neural network (Mask R-CNN) having no region proposal network (Laradji : FIGS. 2-3; Page 6, 3rd paragraph, “simply use the trained Mask R-CNN to predict the object masks for an unseen image. To refine these masks, we leverage the same object proposal method as that used in training.”; A person of ordinary skills in the art would understand that no new bounding box is needed for predicting the 2nd mask with trained Mask R-CNN, thus no region proposal network is not required for this trained Mask R-CNN).
-Regarding claims 4, 11 and 17, Morvan in view Laradji teaches the method of claim 1, the processor of claim 8, and the system of claim 15. The combination further teaches wherein the first segmentation network is trained to generate the first segmentation for at least one object class or at least one pose other than an object class or a pose on which the first segmentation network was trained (Morvan: FIG. 6; [0053], “the NN may be implemented and trained in one of various ways”; FIGS. 2-4; [0052]; Note: the NN may be trained with ground truth data only with portions of distal tibia 402 and distal fibula 404 that are in bounding box 400 and the NN may be required to generate a first segmentation mask for data with remaining portion of distal tibia 402 and distal fibula 404 outside bounding 400; See also Galeev: FIGS. 2-3; [0042], “model that is trained on a given dataset … pre-trained models may be used, such as StyleGAN”; [0046], “The decoder is trained in a supervised manner”; FIGS. 8-9; [0027], “generalize on real data”). Please not that a person of ordinary skills in the art would understand that any supervised machine leaning model is trained with ground truth data set, The trained mode is always tested and validated with different data set, and is used for production with real data (See Alpaydin: chapter 2 Supervised Learning, MIT Press 2009). That is, the first segmentation network has to be trained to generate the first segmentation for at least one object class or at least one pose other than an object class or a pose on which the first segmentation network was trained will be used for testing, validation, and application.
-Regarding claims 5, 12 and 18, Morvan in view of Laradji teaches the method of claim 1, the processor of claim 8, and the system of claim 15.
Morvan does not disclose providing an image as input to the second segmentation network.
In the same field of endeavor, Laradji teaches a method for instance segmentation by obtaining pseudo masks lead to a supervision for Mask R-CNN (Laradji: Abstract; FIGS. 1-5; Sec. 3.). Laradji further teaches providing an image as input to the second segmentation network (Laradji: FIGS. 2-3).
Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to combine the teaching of Morvan with the teaching of Laradji by using a modified instance segmentation model to infer a second segmentation mask based on the loss comparison with the 1st segmentation mask in order to improve performance of segmentation without manually labeling (Laradji: Abstract; Sec. 1., last paragraph).
-Regarding claims 6, 13 and 19, Morvan in view of Laradji teaches the method of claim 5, the processor of claim 12, and the system of claim 18. The combination further teaches wherein at least one of the one or more objects corresponds to a class for which the second segmentation network was not trained (Morvan: FIG. 4, distal tibia 402 and distal fibula 404 outside of bounding box 400 ; [0052]). Note: a person of ordinary skills in the art would understand that any supervised machine leaning model is trained with ground truth data set, The trained mode is always tested and validated with different data set, and is used for production with real data (See Alpaydin: chapter 2 Supervised Learning, MIT Press 2009). That is, at least one of the one or more objects corresponds to a class for which the second segmentation network was not trained will be used for testing, validation, and application.
-Regarding claims 7 and 14, , Morvan in view Laradji teaches the method of claim 1 and the processor of claim 8. The combination further teaches wherein the first segmentation network includes an image encoder for feature extraction and a mask decoder for generating segmentation masks (Morvan: FIG. 6; [0053]; [0074], “The encoder extracts high level features …”,[0076], “decoder”; See Galeev as well: Galeev: FIGS. 1-3; [0009]).
-Regarding claim 20, , Morvan in view Laradji teaches the system of claim 15. The combination further teaches wherein the system comprising: a system for rendering graphical output (Morvan: FIG. 1, display 110; FIG.2, unit 208; [0048]), a system for generating or presenting virtual reality (VR) content, a system for generating or presenting mixed reality (MR) content (Morvan: FIG. 1, visualization device 116; [0034], “visualization device 116 may be a mixed reality (MR) visualization device, virtual reality (VR) visualization device”), or a system for synthetic data generation (Galeev: [0055], “The synthetic dataset may be generated using the trained GAN and the trained decoder in operation S230”; FIG. 3).
Response to Arguments
Applicant’s arguments with respect to claim rejections under 35 U.S.C. 103 for claims 1-20 dated on 04/03/2026 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Zhao et al, Pseudo Mask Augmented Object Detection, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 4061-4070), hereinafter Zhao teaches a method for instance-level segmentation by recursively estimating the pseudo ground-truth object masks from the instance-level object segmentation network training, and then enhancing the detection network with top-down segmentation feedback. The pseudo ground truth mask and network parameters are optimized alternatively to mutually benefit each other.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to XIAO LIU whose telephone number is (571)272-4539. The examiner can normally be reached Monday-Thursday and Alternate Fridays 8:30-4:30.
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, Jennifer Mehmood can be reached at (571) 272-2976. 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.
/XIAO LIU/Primary Examiner, Art Unit 2664
Prosecution Insights