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 .
Claim Status
The status of claims 1-20 is:
Claims 1-20 were restricted as of the Requirement for Restriction mailed 04/15/2026.
Claims 1 and 14-15 are amended as of the remarks and amendments received 06/15/2026.
Claims 2-13 and 16-20 remain as originally presented as of the remarks and amendments received 06/15/2026.
Election/Restrictions
In light of the amendments made 06/15/2026, the election requirement is withdrawn as the claims are no longer patently distinct and there is no longer a serious burden for search or examination as amended claim 1 is now a linking claim.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 07/19/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Claim Interpretation
All claim limitations that include “optionally” are interpreted as not being required for the claim. If Applicant wishes for those claim limitations to binding on the claims, “optionally” should be removed.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefore, subject to the conditions and requirements of this title.
Claims 15-17 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because the BRI of a “computer-readable medium” includes software per se. Examiner suggests amending the claim to state “a non-transitory computer-readable medium”.
Claim Rejections - 35 USC § 102
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.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 4 and 13 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Choi et al. (Choi, J., & Oh, T. H. (2023). Joint video super-resolution and frame interpolation via permutation invariance. Sensors, 23(5), 2529., presented in the IDS received 07/19/2024, hereinafter “Choi”).
Regarding claim 4, Choi discloses a method of training a model for generating super resolution images of an object type, wherein the model comprises a correspondence estimation network (Choi Fig. 1: flow estimation module) and a reconstruction network (Choi Fig. 1: permutation invariant residual network), the method comprising:
obtaining a set of training data for an object of the object type (Choi Page 6: “We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”, videos of faces),
the set of training data comprising a plurality of images of the object at different viewpoints and optical flow fields between pairs of images of the plurality (Choi Page 6: “We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”, videos of faces from all different angles and with different optical flow fields);
training the correspondence estimation network using the set of training data, and obtaining a trained reconstruction network (Choi Page 6: “We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”).
Regarding claim 13, Choi discloses the method, wherein the object type is any one of: license plates; faces (Choi Fig. 4); billboards; signs.
Claim Rejections - 35 USC § 103
Claim(s) 1-2 are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (CN 107480772 A presented in the IDS received 07/19/2024 but using the translation provided herein, hereinafter “Wang”) in view of Choi and Mayer et al. (Mayer, N., Ilg, E., Fischer, P., Hazirbas, C., Cremers, D., Dosovitskiy, A., & Brox, T. (2018). What makes good synthetic training data for learning disparity and optical flow estimation?. International Journal of Computer Vision, 126(9), 942-960., hereinafter “Mayer”).
Regarding claim 1, Wang discloses a method of generating a super resolution image of an object (Wang Abstract: “The invention claims a license plate super-resolution processing based on deep learning method and system”), the method comprising:
receiving a plurality of frames of a video of the object (Wang Abstract: “comprising obtaining a series of images from the original compressed monitoring video comprises license information”);
extracting from the plurality of frames a plurality of images of the object (Wang Page 2: “performing target detection to each frame image, obtaining the target area of each frame of image”);
selecting an image of the plurality of images as a target image (Wang Page 3: “selecting n located at the interest point on the track of interest, and using the interest point as centre captured image, wherein any one image is to be super-resolution frame, the frame to be super-resolution image to register with the other”);
applying a trained model to the plurality of images to generate a super resolution image of the object (Wang Abstract: “The invention claims a license plate super-resolution processing based on deep learning method and system”) wherein the trained model comprises:
(a) a module configured to compute a respective optical flow between the target image and each other image of the plurality (Wang Page 3: “Finally, for registering the multi-frame image capturing, firstly estimating the movement calculating each frame reference frame relative to the movement vector to be super-resolution frame, motion estimation can be motion estimation method by the optical flow method”), and
(b) a reconstruction neural network configured to generate a super resolution version of the target image using the plurality of images and the respective optical flows between the target image and each other image of the plurality (Wang Page 5: “step S1010, using deep network weight to super-resolution processing the multi-frame image after registration, obtaining clear license plate I with high resolution”; Wang Page 5: “step S108, selecting n located at the interest point on the track of interest, and using the interest point as centre captured image, wherein any one image is to be super-resolution frame, the frame to be super-resolution image to register with the other”; Wang Page 3: “Finally, for registering the multi-frame image capturing, firstly estimating the movement calculating each frame reference frame relative to the movement vector to be super-resolution frame, motion estimation can be motion estimation method by the optical flow method”).
Wang does not explicitly disclose the method comprising:
(a) a correspondence estimation neural network configured to compute a respective optical flow between the target image and each other image of the plurality, and trained for the object type of the object; and
wherein the correspondence estimation neural network is trained using a set of training data for an object type of the object, the set of training data comprising the plurality of images of the object at different viewpoints and optical flow fields between pairs of images of the plurality, the images and optical flow fields being generated from a digital 3D model of the object of the object type.
However, Choi teaches (a) a correspondence estimation neural network configured to compute a respective optical flow between the target image and each other image of the plurality, and trained for the object type of the object (Choi Fig. 1: flow estimation module; Choi Page 4: “Given two input frames ILR1 and ILR2 ,the flow estimation CNN estimates the bidirectional flow between them, yielding flow maps F1→2 and F2→1. Then, we backward-warp each frame to the intermediate position by applying half the magnitude of the flow maps, producing warped frames W1 and W2”); and
wherein the correspondence estimation neural network is trained using a set of training data for an object type of the object (Choi Page 6: “We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”, videos of faces), the set of training data comprising the plurality of images of the object at different viewpoints and optical flow fields between pairs of images of the plurality (Choi Page 6: “We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”, videos of faces from all different angles and with different optical flow fields).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the flow estimation module as taught by Choi with the method of Wang because it would improve the method as a network is trainable and will lead to more accurate optical flow calculations (Choi Page 2: “Our approach demonstrates superior performance in terms of quantitative comparisons and visual results”). This motivation for the combination of Wang and Choi is supported by KSR exemplary rationale (G) Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention and rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
The Wang and Choi combination does not explicitly disclose the method comprising:
the images and optical flow fields being generated from a digital 3D model of the object of the object type.
However, Mayer teaches the method comprising:
the images and optical flow fields being generated from a digital 3D model of the object of the object type (Mayer Page 946: “A simple way of creating data with ground truth displacements is to take images of objects (segmented such that anything but the object itself is masked out) and paste them onto a background, with a randomized transformation applied from the first to the second image. Figure 2 illustrates how a FlyingChairs sample is created this way. We used images of chairs created from CAD models by Aubry et al. (2014)”; Mayer Fig. 2).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the 3D model images as taught by Mayer with the method of Wang and Choi because it would improve the method as it would allow for the model to be trained on any type of object and because the data is very easy to create (Mayer Page 943). This motivation for the combination of Wang, Choi, and Mayer is supported by KSR exemplary rationale (G) Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention and rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Regarding claim 2, Wang discloses the method wherein the step of extracting comprises applying a trained object detector to the plurality of frames (Wang Page 6: “method of object detection can also be used R-CNN, Fast R-CNN, Faster R-CNN, SSD, YOLO, etc. based on target detection algorithm of deep learning to detect of interest”), optionally wherein the trained object detector is a You Only Look Once (YOLO) type detector (Wang Page 6: “method of object detection can also be used R-CNN, Fast R-CNN, Faster R-CNN, SSD, YOLO, etc. based on target detection algorithm of deep learning to detect of interest”), further optionally wherein the step of extracting further comprises applying a trained object tracker to identify the same object in each frame from a plurality of objects in one or more frames having a common object type.
Claim(s) 3 is rejected under 35 U.S.C. 103 as being unpatentable over the Wang, Choi, and Mayer combination in view of Teed et al. (Teed, Z., & Deng, J. (2020, August). Raft: Recurrent all-pairs field transforms for optical flow. In European conference on computer vision (pp. 402-419). Cham: Springer International Publishing., presented in the IDS received 07/19/2024, hereinafter “Teed”).
Regarding claim 3, the Wang, Choi, and Mayer combination does not explicitly disclose the method wherein the correspondence estimation network is a Recurrent All-Pairs Field Transforms (RAFT) neural network.
However, Teed teaches the method wherein the correspondence estimation network is a Recurrent All-Pairs Field Transforms (RAFT) neural network (Teed Abstract: “We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state of-the-art performance”).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate RAFT network as taught by Teed with the method of Wang, Choi, and Mayer because it would improve the method the RAFT network is the state of the art network for optical flow calculations (Teed Abstract: “We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state of-the-art performance”; Choi Page 15: “Nonetheless, our method can be easily improved if we replace the optical flow module with a more advanced state-of-the-art optical flow method, e.g., [46,47]”, explicitly citing Teed as a suggested improvement). This motivation for the combination of Wang, Choi, Mayer, and Teed is supported by KSR exemplary rationale (G) Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention and rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Claim(s) 5-6 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Choi in view of Teed.
Regarding claim 5, Choi does not explicitly disclose the method wherein the correspondence estimation network is a Recurrent All-Pairs Field Transforms (RAFT) neural network.
However, Teed teaches the method wherein the correspondence estimation network is a Recurrent All-Pairs Field Transforms (RAFT) neural network (Teed Abstract: “We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state of-the-art performance”).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate RAFT network as taught by Teed with the method of Choi because it would improve the method the RAFT network is the state of the art network for optical flow calculations (Teed Abstract: “We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state of-the-art performance”; Choi Page 15: “Nonetheless, our method can be easily improved if we replace the optical flow module with a more advanced state-of-the-art optical flow method, e.g., [46,47]”, explicitly citing Teed as a suggested improvement). This motivation for the combination of Choi and Teed is supported by KSR exemplary rationale (G) Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention and rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Regarding claim 6, Choi discloses the method wherein the step of training is carried out using a loss function comprising norms of the differences between the optical flows calculated by correspondence estimation network for pairs of images in the set of training data and the corresponding optical flows in the set of training data (Choi Page 6: “To train our network, we utilize subsequent frame triplets provided in high-resolution IHR 1 , IHR 2 ,and IHR 3 ,and down-sample (bicubic) them to low-resolution images. Thus, given the low-resolution images ILR 1 and ILR 3 as input, our model produces the interpolated high resolution frame OHR 2. We use the pixel-wise 1-lossdefinedasL1= IHR 2 −OHR 2 1. We also apply the perceptual loss utilizing the response from the relu4_3 layer of VGG-19 [40]: Lp= φ(IHR 2 )−φ(OHR 2 ) 2 2 , where φ(·) denotes the feature vector of the relu4_3layer. We take the sum of L1 and Lp as the final loss, Ltotal=λL1+µLp, where we set λ and µ to 2.0 and 0.01 respectively”).
Regarding claim 11, Choi discloses the method wherein said training the reconstruction network uses the trained correspondence estimation algorithm to compute respective optical flow fields between pairs of the images in the further set of training data (Choi Page 6: “To train our network, we utilize subsequent frame triplets provided in high-resolution IHR 1 , IHR 2 ,and IHR 3 ,and down-sample (bicubic) them to low-resolution images. Thus, given the low-resolution images ILR 1 and ILR 3 as input, our model produces the interpolated high resolution frame OHR 2. We use the pixel-wise 1-lossdefinedasL1= IHR 2 −OHR 2 1. We also apply the perceptual loss utilizing the response from the relu4_3 layer of VGG-19 [40]: Lp= φ(IHR 2 )−φ(OHR 2 ) 2 2 , where φ(·) denotes the feature vector of the relu4_3layer. We take the sum of L1 and Lp as the final loss, Ltotal=λL1+µLp, where we set λ and µ to 2.0 and 0.01 respectively. . . . We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”).
Claim(s) 8 is rejected under 35 U.S.C. 103 as being unpatentable over the Choi and Teed combination in view of Mayer.
Regarding claim 8, the Choi and Teed combination does not explicitly disclose the method wherein the step of obtaining a set of training data comprises:
generating, from a digital 3D model of the object of the object type, a plurality of images of the object at respective different viewpoints;
and calculating, using the 3D model, for a plurality of pairs of the images, a respective optical flow field between the images of the pair.
However, Mayer teaches the method wherein the step of obtaining a set of training data comprises:
generating, from a digital 3D model of the object of the object type, a plurality of images of the object at respective different viewpoints (Mayer Page 946: “A simple way of creating data with ground truth displacements is to take images of objects (segmented such that anything but the object itself is masked out) and paste them onto a background, with a randomized transformation applied from the first to the second image. Figure 2 illustrates how a FlyingChairs sample is created this way. We used images of chairs created from CAD models by Aubry et al. (2014)”);
and calculating, using the 3D model, for a plurality of pairs of the images, a respective optical flow field between the images of the pair (Mayer Fig. 2 description: “All transforms are affine which makes computing the ground truth flow field easy”, optical flow field shown in outputs).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the 3D model images as taught by Mayer with the method of Choi and Teed because it would improve the method as it would allow for the model to be trained on any type of object and because the data is very easy to create (Mayer Page 943). This motivation for the combination of Choi, Teed, and Mayer is supported by KSR exemplary rationale (G) Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention and rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Claim(s) 9-10 are rejected under 35 U.S.C. 103 as being unpatentable over the Choi and Teed combination in view of Wang.
Regarding claim 9, the Choi and Teed combination does not explicitly disclose the method wherein the step of obtaining a trained reconstruction network comprises:
training the reconstruction network using a further set of training data, the further set of training data comprising one or more images of an object of the object type at the super resolution and a plurality of images of the object at resolutions lower than the super resolution, optionally, wherein the images in the further set of training data are generated from the images in the initial set of training data.
However, Wang teaches the method wherein the step of obtaining a trained reconstruction network comprises:
training the reconstruction network using a further set of training data (Wang Page 5: “step S1012, the extracted character in low-quality license and high resolution license plate, aiming at each character, establishes deep learning training library of n low-quality sample and 1 of high resolution samples, wherein high-resolution license obtained from the original monitoring video, the low-quality license plate is low-quality video compression and different from the original monitoring video resolution down-sampling obtained in the obtaining”), the further set of training data comprising one or more images of an object of the object type at the super resolution and a plurality of images of the object at resolutions lower than the super resolution (Wang Page 5: “step S1012, the extracted character in low-quality license and high resolution license plate, aiming at each character, establishes deep learning training library of n low-quality sample and 1 of high resolution samples, wherein high-resolution license obtained from the original monitoring video, the low-quality license plate is low-quality video compression and different from the original monitoring video resolution down-sampling obtained in the obtaining”), optionally, wherein the images in the further set of training data are generated from the images in the initial set of training data.
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the further set of training data as taught by Wang with the method of Choi and Teed because it would improve the method as providing more training data will allow for more robust training and improve the accuracy of the method. This motivation for the combination of Choi, Teed, and Wang is supported by KSR exemplary rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Regarding claim 10, Choi discloses the method wherein said training the reconstruction network uses the trained correspondence estimation algorithm to compute respective optical flow fields between pairs of the images in the further set of training data (Choi Page 6: “To train our network, we utilize subsequent frame triplets provided in high-resolution IHR 1 , IHR 2 ,and IHR 3 ,and down-sample (bicubic) them to low-resolution images. Thus, given the low-resolution images ILR 1 and ILR 3 as input, our model produces the interpolated high resolution frame OHR 2. We use the pixel-wise 1-lossdefinedasL1= IHR 2 −OHR 2 1. We also apply the perceptual loss utilizing the response from the relu4_3 layer of VGG-19 [40]: Lp= φ(IHR 2 )−φ(OHR 2 ) 2 2 , where φ(·) denotes the feature vector of the relu4_3layer. We take the sum of L1 and Lp as the final loss, Ltotal=λL1+µLp, where we set λ and µ to 2.0 and 0.01 respectively. . . . We train our entire framework using the Vimeo90k dataset[12] of size 448×256 using the Adam optimizer with β1=0.9 and β2=0.999, a learning rate of 0.001, and mini-batch size of 16 samples”).
Claim(s) 12 is rejected under 35 U.S.C. 103 as being unpatentable over Choi in view of Cabellero et al. (U.S. Patent No 10701394 B1, hereinafter “Caballero”).
Regarding claim 12, Choi discloses the method wherein the reconstruction neural network comprises a recurrent back projection network and/or wherein the reconstruction network is trained using a loss function comprising a mean absolute error loss component and an image perceptual loss component (Choi Page 6: “We use the pixel-wise 1-lossdefinedasL1= IHR 2 −OHR 2 1. We also apply the perceptual loss utilizing the response from the relu4_3 layer of VGG-19 [40]: Lp= φ(IHR 2 )−φ(OHR 2 ) 2 2 , where φ(·) denotes the feature vector of the relu4_3layer. We take the sum of L1 and Lp as the final loss, Ltotal=λL1+µLp, where we set λ and µ to 2.0 and 0.01 respectively”).
Choi does not explicitly disclose the method wherein the network is trained using a loss function comprising a mean squared error loss component.
However, Caballero teaches the method wherein the network is trained using a loss function comprising a mean squared error loss component (Caballero Col 23 Lines 37-41: “Given a training set consisting of high resolution image examples I.sub.n.sup.HR, n=1 . . . N, the corresponding low resolution images I.sub.n.sup.LR, n=1 . . . N are generated, and the pixel-wise mean squared error (MSE) of the reconstruction is calculated as an objective function to train the network”).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the mean squared error as taught by Caballero with the method of Choi because it would penalize larger errors more. This motivation for the combination of Choi and Caballero is supported by KSR exemplary rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Claim(s) 14-16 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over the Wang, Choi, and Mayer combination in view of Cabellero.
Regarding claim 14, it is rejected under the same analysis as claim 1 above along with Caballero’s teaching of a processor and computer-readable medium storing instructions that are operable (Cabellero Col 51 lines 6-11: “When implemented in software, firmware, middleware or microcode, the program code or code segments to perform the necessary tasks may be stored in a machine or computer readable medium such as a storage medium. A processor(s) may perform the necessary tasks”).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the processor and computer-readable medium as taught by Caballero with the method of Wang Choi, and Mayer because a processor and computer-readable medium are necessary for the method to be useable on a computer or other processing device. This motivation for the combination of Wang, Choi, Mayer, and Caballero is supported by KSR exemplary rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Regarding claim 15, it is rejected under the same analysis as claim 1 above along with Caballero’s teaching of a computer-readable medium storing a computer program (Caballero Col. 8 lines 50-53: “In a general aspect, a method and a non-transitory computer readable medium (that includes code segments that when executed by a processor cause the processor to)”).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the computer readable medium as taught by Caballero with the method of Wang Choi, and Mayer because it would allow the method to be run on any processing device. This motivation for the combination of Wang, Choi, Mayer, and Caballero is supported by KSR exemplary rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Regarding claim 16, it is rejected under the same analysis as claim 2 above.
Regarding claim 18, it is rejected under the same analysis as claim 2 above.
Claim(s) 17 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over the Wang, Choi, Mayer, and Caballero combination in view of Teed.
Regarding claim 17, the Wang, Choi, Mayer, and Caballero combination does not explicitly disclose the computer-readable medium wherein the correspondence estimation network is a Recurrent All-Pairs Field Transforms (RAFT) neural network.
However, Teed teaches the computer-readable medium wherein the correspondence estimation network is a Recurrent All-Pairs Field Transforms (RAFT) neural network (Teed Abstract: “We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state of-the-art performance”).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to incorporate RAFT network as taught by Teed with the computer-readable medium of Wang, Choi, and Mayer because it would improve the medium the RAFT network is the state of the art network for optical flow calculations (Teed Abstract: “We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state of-the-art performance”; Choi Page 15: “Nonetheless, our method can be easily improved if we replace the optical flow module with a more advanced state-of-the-art optical flow method, e.g., [46,47]”, explicitly citing Teed as a suggested improvement). This motivation for the combination of Wang, Choi, Mayer, Caballero, and Teed is supported by KSR exemplary rationale (G) Some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention and rationale (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results.
Regarding claim 19, it is rejected under the same analysis as claim 17 above.
Regarding claim 20, it is rejected under the same analysis as claim 3 above.
Allowable Subject Matter
Claim 7 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AIDAN KEUP whose telephone number is (703)756-4578. The examiner can normally be reached Monday - Friday 8:00-4:00.
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.
/AIDAN KEUP/ Examiner, Art Unit 2666
/Molly Wilburn/Primary Examiner, Art Unit 2666
Prosecution Insights