Prosecution Insights
Last updated: April 19, 2026
Application No. 18/460,053

SYSTEMS AND METHODS FOR AN ADAPTIVE AND REGION-SCALE PROPOSING MECHANISM FOR OBJECT RECOGNITION SYSTEMS

Final Rejection §103
Filed
Sep 01, 2023
Examiner
LIU, XIAO
Art Unit
2664
Tech Center
2600 — Communications
Assignee
Arizona Board of Regents
OA Round
2 (Final)
89%
Grant Probability
Favorable
3-4
OA Rounds
2y 9m
To Grant
99%
With Interview

Examiner Intelligence

Grants 89% — above average
89%
Career Allow Rate
257 granted / 290 resolved
+26.6% vs TC avg
Moderate +12% lift
Without
With
+11.5%
Interview Lift
resolved cases with interview
Typical timeline
2y 9m
Avg Prosecution
44 currently pending
Career history
334
Total Applications
across all art units

Statute-Specific Performance

§101
8.8%
-31.2% vs TC avg
§103
50.9%
+10.9% vs TC avg
§102
17.0%
-23.0% vs TC avg
§112
17.4%
-22.6% vs TC avg
Black line = Tech Center average estimate • Based on career data from 290 resolved cases

Office Action

§103
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 Applicant’s amendments filed on 11/24/2025 to the specification and claims have overcome specification objection and claim rejections under 35 U.S.C. 112(b) as preciously set forth in the Non-Final Rejection Office Action mailed on 08/28/2025. 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, 7, 12-15 and 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato. -Regarding claim 1, Mullapudi discloses a system for adaptive object recognition with reduced computational costs, comprising: a processor in communication with a memory, the memory including instructions, which, when executed (one or more memories and processors has to be used in order to implement Mullapudi’s FIGS. 1-2, Algorithm 1), cause the processor to (Abstract; FIGS. 1-4 PNG media_image1.png 282 373 media_image1.png Greyscale ): access input frames including images associated with vehicular traffic generated by a camera (p. 3573, 1st Col., 1st paragraph; 2nd Col., 1st paragraph; FIG. 1); and apply a first neural network implemented as a region-of-interest proposal module to identify a plurality of regions of interest (ROls) in the input frames that satisfy a predetermined probability of object existence (FIG. 1, Mask RCNN; Algorithm 1), the first neural network trained to generate a binary segmentation mask that distinguishes foreground regions from background regions in the input frames (FIG. 1, Mask RCNN; Algorithm 1; p. 3574, 2nd Col., last paragraph; It is well-known that Mask RCNN adopts two-stage procedure with Regional Proposal Network as a first stage and in parallel to predicting the class and box offset as a second stage, and Mask RCNN is trained using multi-task loss on each sampled ROI. Mask R-CNN also outputs a binary mask for each ROI. See He et al (2017 ICVV)) wherein the ROls define specific portions (plain meaning of ROI) of the input frames on which a second neural network is configured to focus (FIG. 1, JITNet; FIG. 2; Algorithm 1; p. 3575, Sec. 3.2), the second neural network being applied to the ROls identified by the first neural network to reduce computational load for object recognition (FIG. 1, caption, “low-cost student model is tasked to generate a high-resolution, per-frame semantic segmentation. To retain high accuracy, as new frames arrive, an expensive teacher model’s (MRCNN) output is periodically used as a learning target to adapt the student and selecting the next frame to request supervision”; Algorithm 1; p. 3573, 2nd Col., 1st paragraph, “a lightweight “student “ model output the predictions of a larger, reliable high-capacity “teacher”, 2n paragraph). Mullapudi does not disclose a binary neural network to generate a binary segmentation mask. In the same field of endeavor, Sato teaches a mask structure optimization method (Sato: Abstract; FIGS. 1-20). Sato further teaches a binary neural network to generate a binary segmentation mask (Sato: [0014], “the mask pattern optimization unit may optimize the mask pattern of the mask using a binary convolutional neural network”; [0052], “ a binary mask”; [0120]; FIGS. 1, 10). 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 Mullapudi with the teaching of Sato by comprising a binary convolutional neural network in the first neural network in order optimize mask pattern of the mask (Sato: [0014]-[0015]) and reduce computation cost. -Regarding claim 19, Mullapudi discloses a method of adaptive object recognition with reduced computational costs, comprising (Abstract; FIGS. 1-4): accessing input frames including images associated with vehicular traffic generated by a camera (p. 3573, 1st Col., 1st paragraph; 2nd Col., 1st paragraph; FIG. 1); and applying a first neural network implemented as a region-of-interest proposal module to identify a plurality of regions of interest (ROls) in the input frames that satisfy a predetermined probability of object existence (FIG. 1, Mask RCNN; Algorithm 1; ), the first neural network trained to generate a binary segmentation mask that distinguishes foreground regions from background regions in the input frames (FIG. 1, Mask RCNN; Algorithm 1; p. 3574, 2nd Col., last paragraph; It is well-known that Mask RCNN adopts two-stage procedure with Regional Proposal Network as a first stage and in parallel to predicting the class and box offset as a second stage, and Mask RCNN is trained using multi-task loss on each sampled ROI. Mask R-CNN also outputs a binary mask for each ROI. See He et al (2017 ICVV)) wherein the ROls define specific portions (plain meaning of ROI) of the input frames on which a second neural network is configured to focus (FIG. 1, JITNet; FIG. 2; Algorithm 1; p. 3575, Sec. 3.2), the second neural network being applied to the ROls identified by the first neural network to reduce computational load for object recognition (FIG. 1, caption, “low-cost student model is tasked to generate a high-resolution, per-frame semantic segmentation. To retain high accuracy, as new frames arrive, an expensive teacher model’s (MRCNN) output is periodically used as a learning target to adapt the student and selecting the next frame to request supervision”; Algorithm 1; p. 3573, 2nd Col., 1st paragraph, “a lightweight “student “ model output the predictions of a larger, reliable high-capacity “teacher”, 2n paragraph). Mullapudi does not disclose a binary neural network to generate a binary segmentation mask. In the same field of endeavor, Sato teaches a mask structure optimization method (Sato: Abstract; FIGS. 1-20). Sato further teaches a binary neural network to generate a binary segmentation mask (Sato: [0014], “the mask pattern optimization unit may optimize the mask pattern of the mask using a binary convolutional neural network”; [0052], “ a binary mask”; [0120]; FIGS. 1, 10). 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 Mullapudi with the teaching of Sato by comprising a binary convolutional neural network in the first neural network in order optimize mask pattern of the mask (Sato: [0014]-[0015]) and reduce computation cost. -Regarding claim 20, Mullapudi discloses a non-transitory, computer-readable medium storing instructions encoded thereon, the instructions, when executed by one or more processors (one or more memories and processors has to be used in order to implement Mullapudi’s FIGS. 1-2, Algorithm 1), cause the one or more processors to perform operations to (Abstract; FIGS. 1-4): access input frames including images associated with vehicular traffic generated by a camera (p. 3573, 1st Col., 1st paragraph; 2nd Col., 1st paragraph; FIG. 1); and apply a first neural network implemented as a region-of-interest proposal module to identify a plurality of regions of interest (ROls) in the input frames that satisfy a predetermined probability of object existence (FIG. 1, Mask RCNN; Algorithm 1), the first neural network trained to generate a binary segmentation mask that distinguishes foreground regions from background regions in the input frames (FIG. 1, Mask RCNN; Algorithm 1; p. 3574, 2nd Col., last paragraph; It is well-known that Mask RCNN adopts two-stage procedure with Regional Proposal Network as a first stage and in parallel to predicting the class and box offset as a second stage, and Mask RCNN is trained using multi-task loss on each sampled ROI. Mask R-CNN also outputs a binary mask for each ROI. See He et al (2017 ICVV)) wherein the ROls define specific portions (plain meaning of ROI) of the input frames on which a second neural network is configured to focus (FIG. 1, JITNet; FIG. 2; Algorithm 1; p. 3575, Sec. 3.2), the second neural network being applied to the ROls identified by the first neural network to reduce computational load for object recognition (FIG. 1, caption, “low-cost student model is tasked to generate a high-resolution, per-frame semantic segmentation. To retain high accuracy, as new frames arrive, an expensive teacher model’s (MRCNN) output is periodically used as a learning target to adapt the student and selecting the next frame to request supervision”; Algorithm 1; p. 3573, 2nd Col., 1st paragraph, “a lightweight “student “ model output the predictions of a larger, reliable high-capacity “teacher”, 2n paragraph). Mullapudi does not disclose a binary neural network to generate a binary segmentation mask. In the same field of endeavor, Sato teaches a mask structure optimization method (Sato: Abstract; FIGS. 1-20). Sato further teaches a binary neural network to generate a binary segmentation mask (Sato: [0014], “the mask pattern optimization unit may optimize the mask pattern of the mask using a binary convolutional neural network”; [0052], “ a binary mask”; [0120]; FIGS. 1, 10). 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 Mullapudi with the teaching of Sato by comprising a binary convolutional neural network in the first neural network in order optimize mask pattern of the mask (Sato: [0014]-[0015]) and reduce computation cost. -Regarding claim 7, Mullapudi in view of Sato teaches the system of claim 1. The combination further teaches wherein the processor is configured for traffic data collection (Mullapudi: FIG. 1, 3; p. 3576, Sec. 4., 1st paragraph, “traffic violation monitoring”; Tables 2-3). -Regarding claim 12, Mullapudi in view of Sato teaches the system of claim 1. The combination further teaches wherein the first neural network is a light detector of a knowledge distillation model configured for initial object detection and proposal of the ROIs (Mullapudi: caption, “low-cost student model is tasked to generate a high-resolution, per-frame semantic segmentation. To retain high accuracy, as new frames arrive, an expensive teacher model’s (MRCNN) output is periodically used as a learning target to adapt the student and selecting the next frame to request supervision”; Algorithm 1; p. 3573, 2nd Col., 1st paragraph, “a lightweight “student “ model output the predictions of a larger, reliable high-capacity “teacher”, 2n paragraph). -Regarding claim 13, Mullapudi in view of Sato teaches the system of claim 1. The combination further teaches wherein as the light detector, the first neural network outputs objects with high confidence considered accurate detections (Mullapudi: caption, “low-cost student model is tasked to generate a high-resolution, per-frame semantic segmentation. To retain high accuracy, as new frames arrive, an expensive teacher model’s (MRCNN) output is periodically used as a learning target to adapt the student and selecting the next frame to request supervision”; Algorithm 1; p. 3573, 2nd Col., 1st paragraph, “a lightweight “student “ model output the predictions of a larger, reliable high-capacity “teacher”, 2n paragraph). -Regarding claim 14, Mullapudi in view of Sato teaches the system of claim 1. The combination further teaches wherein the first neural network selects the ROIs in which an event was observed (Mullapudi: p. 3579, 2nd Col., 3rd paragraph, “requiring … updates … correspond to events when new birds appear”; FIGS. 1, 4; Algorithm 1). -Regarding claim 15, Mullapudi in view of Sato teaches the system of claim 12. The combination further teaches wherein the input frames are decomposed to separate the ROIs (Mullapudi: FIGS. 1, 3). Claim(s) 2-3 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato, and further in view of Han et al (arXiv:2102.04906v4 2 Dec. 2021), hereinafter Han. -Regarding claim 2, Mullapudi in view of Sato teaches the system of claim 1. Mullapudi in view of Sato does not teaching reducing an input size for the second neural network. However, Han is an analogous art pertinent to the problem to be solved in this application and conducts a survey for dynamic neural networks (Han: Abstract; FIGS. 1-9). Han further teaches reducing an input size for the second neural network (Han: FIG. 8; Page 9, Sec.3.2., 1st paragraph, “performing adaptive inference on regions/patches of input images”). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Han by using focus on specific portions of the input frames associated with the ROIs with second neural network in order to further reduce computational load and improve detection accuracy for object recognition (Han: Page 8, Sec.3., 1st paragraph; Page 9, Sec. 3.2., 1st paragraph). -Regarding claim 3, Mullapudi in view of Sato teaches the system of claim 1. The combination teaches a field of view of a fixed camera in a vision- based intersection management application (Mullapudi: p. 3576, Sec. 4., 2nd paragraph, “fixed view cameras … traffic cameras”; p. 3579, 1st Col., 1st paragraph; Tables 2-3). Mullapudi in view of Sato does not teach wherein the first neural network is configured to decompose the images of the input frames into sub-regions. However, Han is an analogous art pertinent to the problem to be solved in this application and conducts a survey for dynamic neural networks (Han: Abstract; FIGS. 1-9). Han further teaches wherein the first neural network is configured to decompose the images of the input frames into sub-regions (Han: Page 9, 1st Col., 1st paragraph, “generates a segmentation mask for an input image, dividing it into m regions”). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Han by using focus on specific portions of the input frames associated with the ROIs with second neural network in order to further reduce computational load and improve detection accuracy for object recognition (Han: Page 8, Sec.3., 1st paragraph; Page 9, Sec. 3.2., 1st paragraph). -Regarding claim 18, Mullapudi in view of Sato teaches the system of claim 12. Mullapudi in view of Sato does not teach decomposing the input frames into independent sub-regions. However, Han is an analogous art pertinent to the problem to be solved in this application and conducts a survey for dynamic neural networks (Han: Abstract; FIGS. 1-9). Han further teaches decomposing the input frames into independent sub-regions (Han: Page 9, 1st Col., 1st paragraph, “generates a segmentation mask for an input image, dividing it into m regions”). 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 Mullapudi in view of Sato with the teaching of Han by using focus on specific portions of the input frames associated with the ROIs with second neural network in order to further reduce computational load and improve detection accuracy for object recognition (Han: Page 8, Sec.3., 1st paragraph; Page 9, Sec. 3.2., 1st paragraph). Claim(s) 4-6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato, and further in view of Deng et al (CN 207993243 U). -Regarding claim 4, Mullapudi in view of Sato teaches the system of claim 1. Mullapudi in view of Sato does not teach a power management assembly positioned proximate to the processor, the power management assembly including a battery in electrical communication with one or more solar panels that powers the processor. However, Deng is an analogous art pertinent to the problem to be solved in this application and teaches a traffic tower implementation (Deng: Abstract; FIGS. 1-2). Deng further teaches a power management assembly positioned proximate to the processor, the power management assembly including a battery in electrical communication with one or more solar panels that powers the processor (Deng: Abstract; FIGS. 1-2). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Deng by using a power management assembly positioned proximate to the processor, the power management assembly including a battery in electrical communication with one or more solar panels that powers the processor to implement the first neural network and the second neural network in order to provide a real-world application for traffic management. -Regarding claim 5, Mullapudi in view of Sato teaches the system of claim 4. Mullapudi in view of Sato teaches reducing power consumption (Mullapudi: p. 3574, 2nd Col., 4th paragraph, “reduce computation”; p. 3575, Sec. 3.1., 2nd paragraph). Mullapudi in view of Sato does not teach using the processor and cameras with a power supply including a battery in electrical communication with one or more solar panels that powers the processor. However, Deng is an analogous art pertinent to the problem to be solved in this application and teaches a traffic tower implementation (Deng: Abstract; FIGS. 1-2). Deng further teaches a power management assembly positioned proximate to the processor, the power management assembly including a battery in electrical communication with one or more solar panels that powers the processor (Deng: Abstract; FIGS. 1-2). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Deng by using a power management assembly positioned proximate to the processor, the power management assembly including a battery in electrical communication with one or more solar panels that powers the processor to implement the first neural network and the second neural network in order to provide a real-world application for traffic management. -Regarding claim 6, Mullapudi in view of Sato teaches the system of claim 4. Mullapudi in view of Sato does not teach wherein the processor and the power management assembly are installed onto an existing traffic light pole or lamp post. However, Deng is an analogous art pertinent to the problem to be solved in this application and teaches a traffic tower implementation (Deng: Abstract; FIGS. 1-2). Deng further teaches wherein the processor and the power management assembly are installed onto an existing traffic light pole or lamp post (Deng: Abstract; FIGS. 1-2). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Deng by using a power management assembly positioned proximate to the processor, the power management assembly including a battery in electrical communication with one or more solar panels that powers the processor to implement the first neural network and the second neural network in order to provide a real-world application for traffic management. Claim(s) 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato, and further in view of Wang et al (US 20230222325 A1), hereinafter Wang. -Regarding claim 8, Mullapudi in view of Sato teaches the system of claim 1. The combination further teaches wherein the first neural network reduces computational load in a vision-based intersection management application in which the cameras are fixed and a domain of an observing environment does not change, thereby allowing temporal knowledge adaptation and implementation of the first neural network as a light model at inference time (Mullapudi: FIGS. 1; p. 3576, Sec. 4., 2nd paragraph, “fixed view cameras … traffic cameras”; p. 3579, 1st Col., 1st paragraph; Tables 2-3). Mullapudi in view of Sato does not teach wherein the first neural network is a binary neural network defining different quantization structures. However, Wang is an analogous art pertinent to the problem to be solved in this application and teaches a binary neural network model training method constructing an online distillation-enhanced binary neural network training framework (Wang: Abstract; FIGS. 1-3). Wang further teaches wherein the first neural network is a binary neural network defining different quantization structures (Wang: FIG. 1, S100; FIG. 2; [0067], “binarized … quantization”; [0068]). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Wang by using a binary neural network defining different quantization structures for the first neural network in order to reduce computational load. Claim(s) 9-10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato, and further in view of Wang et al (US 20230222325 A1), hereinafter Wang, in view of Qin et al (arXiv:2004.03333v1 31 Mar 2020), hereinafter Qin. -Regarding claim 9, Mullapudi in view of Sato, and further in view of Wang teaches the system of claim 8. Mullapudi in view of Sato, and further in view of Wang does not teach wherein the binary neural network uses binary weights and 4-bit activations. However, Qin is an analogous art pertinent to the problem to be solved in this application and teaches a survey of binary neural network (Qin: Abstract; FIGS. 1-4; Tables 1-7). Qin further teaches wherein the binary neural network uses binary weights and 4-bit activations (Qin: Table 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 modify the teaching of Mullapudi in view of Sato, and further in view of Wang with the teaching of Qin by using binary weights and 4-bit activations in order to optimize the binary neural network for reducing computation load. -Regarding claim 10, Mullapudi in view of Sato, and further in view of Wang teaches the system of claim 8. Mullapudi in view of Sato, and further in view of Hoang does not teach wherein the binary neural network uses binary weights and 4-bit activations. However, Qin is an analogous art pertinent to the problem to be solved in this application and teaches a survey of binary neural network (Qin: Abstract; FIGS. 1-4; Tables 1-7). Qin further teaches wherein the binary neural network uses binary weights and 4-bit activations (Qin: Table 4). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato, and further in view of Wang with the teaching of Qin by using binary weights and 4-bit activations in order to optimize the binary neural network for reducing computation load. Claim(s) 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato, and further in view of Wang et al (US 20230222325 A1), hereinafter Wang, in view of Zhao et al (US 20200012895 A1), hereinafter Zhao. -Regarding claim 11, Mullapudi in view of Sato, and further in view of Wang teaches the system of claim 8. The modification further teaches a segmentation mask in which foreground regions correspond to target objects within a field of view of a fixed traffic or street light pole camera and background regions correspond to remaining portions of the input frames to accommodate selection of the ROls for various objects that occupy reduced portions of a whole frame of the input frames (Mullapudi: FIGS. 1; p. 3576, Sec. 4., 2nd paragraph, “fixed view cameras … traffic cameras”; p. 3579, 1st Col., 1st paragraph; Tables 2-3; FIG. 3). Mullapudi in view of Sato, and further in view of Wang does not teach wherein the binary neural network generates a binary segmentation mask for single class segmentation. However, Zhao is an analogous art pertinent to the problem to be solved in this application and teaches a method for classification and localization (Zhao: Abstract; FIGS. 1-10). Zhao further teaches wherein the binary neural network generates a binary segmentation mask for single class segmentation (Zhao: [0025], “a binary neural network architecture that performs machine learning associated with one or more binary classifications for the medical imaging data”; [0026], “plurality of masks”; [0038]; [0037], “regions of interest”; [0051]; FIG. 4; FIG. 8, step 804). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato, and further in view of Wang with the teaching of Zhao by using the binary neural network to generate a binary segmentation mask for single class segmentation in order to accommodate selection of the ROIs for various objects and reduce computations by the second neural network by reducing the processing regions. Claim(s) 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato, and further in view of Urtasun et al (US 20190146497 A1), hereinafter Urtasun. -Regarding claim 16, Mullapudi in view of Sato teaches the system of claim 1. Mullapudi in view of Sato does not teach a gather-scatter approach that gathers the ROIs into a single dense matrix prior to application of convolution operations. It is well-known that a sparse neural network may be used by applying pruning to a dense neural network. Urtasun is an analogous art pertinent to the problem to be solved in this application and teaches sparse convolutional neural networks (Urtasun: Abstract; FIGS. 1-8). Urtasun further teaches a gather-scatter approach that gathers the ROIs into a single dense matrix prior to application of convolution operations (Urtasun: FIGS. 4, 6; [0054], “portions of the imagery for which the corresponding predictions have low confidence can be included in the region of interest while portions of the imagery”; [0155]; [0160]). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Urtasun by using sparse convolutional neural network with gather-scatter approach that gathers the ROIs into a single dense matrix prior to application of convolution operations in order to gain inference speedup by skipping performing convolutions over regions of the imagery where the imagery is sparse and/or regions of the imagery that are not relevant to the prediction being sought. Claim(s) 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Mullapudi et al (arXiv:1812.02699 2018), hereinafter Mullapudi in view of Sato et al (US 20210166094 A1), hereinafter Sato and further in view of Mao et al (2017 IEEE Third International Conference on Big Data Computing Service and Applications), hereinafter Mao. -Regarding claim 17, Mullapudi in view of Sato teaches the system of claim 1. Mullapudi in view of Sato does not teach implement bin-packing to put the regions (Rols) next to each other using a Maximal rectangles best short side fit approach. However, Mao is an analogous art pertinent to the problem to be solved in this application and teaches a deep learning approach to optimized variable sized bin packing (Mao: Abstract; Page 81, Sec. II. Bin Pack). Mao further teaches implement bin-packing to put the regions (Rols) next to each other using a Maximal rectangles best short side fit approach (Mao: Page 80, Sec. I., 1st paragraph, “pack a set of boxes into a set of boxes”; Page 81, Sec. III. Heuristic and Optimization Decision). Therefore, it would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to modify the teaching of Mullapudi in view of Sato with the teaching of Mao by implementing bin-packing to put the regions (Rols) next to each other using heuristic algorithm in order to efficient process regions of interest. Response to Arguments Applicant’s arguments with respect to claim(s) 1-20 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 Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a). A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action. 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
Read full office action

Prosecution Timeline

Sep 01, 2023
Application Filed
Aug 26, 2025
Non-Final Rejection — §103
Nov 24, 2025
Response Filed
Jan 26, 2026
Final Rejection — §103 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12603972
WIRELESS TRANSMITTER IDENTIFICATION IN VISUAL SCENES
2y 5m to grant Granted Apr 14, 2026
Patent 12592069
OBJECT RECOGNITION METHOD AND APPARATUS, AND DEVICE AND MEDIUM
2y 5m to grant Granted Mar 31, 2026
Patent 12579834
Information Extraction Method and Apparatus for Text With Layout
2y 5m to grant Granted Mar 17, 2026
Patent 12576873
SYSTEM AND METHOD OF CAPTIONS FOR TRIGGERS
2y 5m to grant Granted Mar 17, 2026
Patent 12573175
TARGET TRACKING METHOD, TARGET TRACKING SYSTEM AND ELECTRONIC DEVICE
2y 5m to grant Granted Mar 10, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

AI Strategy Recommendation

Get an AI-powered prosecution strategy using examiner precedents, rejection analysis, and claim mapping.
Powered by AI — typically takes 5-10 seconds

Prosecution Projections

3-4
Expected OA Rounds
89%
Grant Probability
99%
With Interview (+11.5%)
2y 9m
Median Time to Grant
Moderate
PTA Risk
Based on 290 resolved cases by this examiner. Grant probability derived from career allow rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month