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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 11/10/2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Amendment
The amendment and written response filed have been entered and considered
Claims 1, 7, 8, 9, 10, and 11 are currently amended
Claim 6 is cancelled
Claims 12-18 are new
Claim Rejections - 35 USC § 112
Claim 11 is rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. The newly added limitation of amended claim 11 reciting “wherein the first deep learning network learns by using transfer learning from a second deep learning network which extracts feature maps from a left image and a right image to generate a disparity map.” There is no support in the specification In regards to a data generation method that estimates a depth map. Paragraph [0056] is the clearest support for original claim but does not disclose newly amended claim limitation.
Claim Rejections - 35 USC § 102
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 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.
Claims 1 and 10 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Jeon et al (Jeon hereinafter “Efficient Multi-Scale Stereo-Matching Network Using Adaptive Cost Volume Filtering”)
As per claim 1
Jeon teaches extracting feature maps from a left image and a right image by using a first deep learning network (Figure 1. 3.1 Feature Extractor “Given a rectified stereo image pair 𝐼𝑙∈ℝ𝑊×𝐻×3 and 𝐼𝑟∈ℝ𝑊×𝐻×3, each image is inserted into a U-Net… extracted from the different positions of the decoder of the feature extractor network) calculating a disparity map by matching the extracted feature maps (Figure 2, Proposed Method: “ Using these feature maps as inputs, a cost aggregation module aggregates the matching costs and generates multi-scale cost volumes… , an initial disparity map is regressed with the softargmax function from the cost volume and hierarchically upsampled and refined with a refinement network [13] to produce a final disparity map) generating a depth map from the disparity map (Introduction: “Using disparity d, camera focal length f, and distance between cameras B, the depth of the pixel can be calculated as 𝑓/𝐵𝑑” Conclusion “We believe that our method sheds light on developing a practical stereo-matching network that generates accurate depth information in real time. In addition, our method can be adopted in a variety of applications that require real-time 3D depth information.”) wherein the first deep learning network learns by using transfer learning from a second deep learning network which extracts feature maps from a left image and a right image to generate a disparity map. (Figure 2 and Figure 8 )
As per claim 10
Claim 10 is the parallel system claim of method claim 1 and will be rejected under the same premise. Jeon’s concepts necessarily use a computerized environment that uses a processor to function such as the “NVIDIA GeForce RTX GPU “ they use in 4.2 Implementation details.
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.
Claims 2-5 and 12-15 are rejected under 35 U.S.C. 103 as being unpatentable over Jeon et al (Jeon hereinafter “Efficient Multi-Scale Stereo-Matching Network Using Adaptive Cost Volume Filtering”) in view of Yang et al (Yang hereinafter “DrivingStereo: A Large-Scale Dataset for Stereo Matching in Autonomous Driving Scenarios”
As per claim 2
Jeon teaches all claim limitations rejected in claim 1’s 102 rejection. See claim 1’s 102 rejection.
Jeon teaches training the stereo matching network using ground truth disparity maps, as evidenced by the disclosed loss function defined between the predicted disparity and the ground truth disparity which necessarily requires the use of disparity maps as training labels (see 3.5 and 3.5.1)
Jeon does not teach wherein the first deep learning network is trained with a training dataset which is comprised of a left image, a right image, and a disparity map which is specifically generated by using a 3D sensor.
Yang teaches training dataset comprises a left image and a right image and a disparity map (Introduction “We construct a large-scale stereo dataset that comprises over 180K images covering a diverse range of driving scenarios, along with a model-guided filtering strategy for producing high-quality disparity labels.” Data acquisition “Among the raw frames, we choose 18,2188 frames to construct our DrivingStereo dataset, where 174,437 frames from 38 sequences are used as training dataset. The remaining 7,751 frames from 4 sequences are manually selected as testing dataset that has a higher quality of disparity labels” ) and the disparity map which is generated by a 3D sensor (“the disparity labels are projected from Li DAR points and filtered by model-guided strategy as well.”) and a deep learning network trained on the dataset (Introduction: “In order to demonstrate the generalization ability of our dataset and metrics, deep stereo models are trained on FlyingThings3D [20], Cityscapes [7], and our DrivingStereo dataset respectively then compared.” Section 5.2 “The stereo 2015 dataset [21] releases 200 pairs of disparity labels along with their foreground masks, in which 142 images have origin index in raw sequences [11]. Here, we select these 142 images as a validation set and sample other 8,260 frames from the raw dataset as a training set”)
Accordingly, it would have been obvious to one of ordinary skill in the art at the time of the invention to train the stereo depth estimation network of Jeon using disparity labels generated from LiDAR sensor data as taught by Yang. The instant application itself recognizes the benefit of incorporating Lidar derived geometric information into stereo based depth estimation. are known to be less reliable. Jeon explicitly relies on supervised learning where disparity predictions are optimized against ground truth disparity through their loss function and the quality of this ground truth directly controls the learning of marching costs in the cost volume. Yang teaches that disparity labels can be generated from Velodyne LiDAR points, providing geometrically accurate disparity supervision. A person of ordinary skill in the art would have been motivated to use LiDAR derived disparity labels in Jeon because these labels improve the correctness of correspondence learning within the cost volume and take away ambiguity in regions where stereo matching can be unreliable. This enhances disparity estimation performance in a predictable manner. Yang defines a superior/known and established version of the exact supervision signal that Jeon’s architecture is designed to learn from.
As per claim 3
Jeon and Yang teach all claim limitations previously rejected in claim 2’s 103 rejection. See claim 2’s 103 rejection
Jeon teaches the first deep learning network comprises a 1-1 deep learning network for extracting feature maps from the left image, and a 1-2 deep learning network for extracting feature maps from the right image. (Figure 2’s feature extractor) Essentially Jeon teaches applying a feature extraction network to each image of the stereo pair thereby providing separate feature extraction processing for the left and right images. This corresponds to extracting feature maps from the left image and the right image using respective processing paths within the first deep learning network. It is well known in the art that stereo matching architecture process left and right images through parallel branches of a neural network, whether implemented as separate networks or as shared weight architecture.
Accordingly, it would have been obvious to one of ordinary skill in the art at the time of the invention to implement feature extraction stage of the stereo matching network of Jeon using respective deep learning networks for left and right images. Jeon processes each image of the stereo pair to generate feature maps for subsequent cost volume construction. Providing separate processing paths preserves view specific features prior to correspondent matching. A person of ordinary skill in the art would have recognized that parallel feature extraction improves reliability of disparity estimation particularly when trained using accurate disparity supervision such as LiDAR derived labels as taught by Yang.
As per claim 4
Jeon and Yang teach all claim limitations previously rejected in claim 2’s 103 rejection. See claim 2’s 103 rejection
Yang teaches that the 3D sensor is a LiDAR sensor (3. Data Construction “In this section, we first introduce the data acquisition system, where stereo images and LiDAR point clouds are simultaneously collected.” 3.14 Data acquisition “The baseline distance between such stereo pair is 54cm, and the field of view (FOV) is 50◦. The LiDAR is also equipped behind the center camera, and the navigation unit is at the rear.”)
Accordingly, a person of ordinary skill in the art at the time this invention was effectively filed would have found it obvious to utilize LiDAR sensor data as the source of disparity labels in the training process of Jeon as taught by Yang. Jeon relies on ground truth disparity to train the cost volumes aggregation and disparity regression components and Yang teaches that those disparity labels can be generated from Velodyne LiDAR points which provide accurate geometric measurements aligned with stereo image pairs. A person of ordinary skill in the art would have been motivated to use LiDAR derived disparity in Jeon to improve the reliability of correspondence learning and disparity estimation by providing geometrically precise supervision.
As per claim 5
Jeon and Yang teach all claim limitations previously rejected in claim 4’s 103 rejection. See claim 4’s 103 rejection.
Yang teaches the disparity map constituting the training dataset is a map that is converted from a depth map generated through the LiDAR sensor. (Yang uses a dataset from KITTI 2015 “where the disparity labels are transformed from Velodyne LiDAR points.” LiDAR measures depth information and the transformation from LiDAR points to disparity necessarily involves converting depth information into disparity value. Therefore, Yang teaches that the disparity map constituting the training dataset in the Jeon/Yang pipeline is converted from a depth map generated through a LiDAR sensor)
Accordingly, a person of ordinary skill in the art at the time this invention was effectively filed would have found it obvious to generate the disparity labels used in the training of Jeon by converting depth information obtained from LiDAR sensor as taught by Yang. Jeon relies on ground truth disparity to train its cost volume aggregation and disparity regression components while Yang teaches that disparity labels in a dataset are obtained by transforming Velodyne LiDAR point data. Since LiDAR inherently provides depth measurements, a person of ordinary skill in the art would have recognized that converting LiDAR derived depth into disparity values in the image domain enables direct use of such data within Jeon’s stereo matching framework. This provides geometric consistent supervision aligned with network’s disparity output.
As per claim 12
Jeon teaches all claim limitations previously rejected in claim 10’s 102 rejection. See claim 10’s 102 rejection.
Jeon’s method and its concepts necessarily use a computerized environment that uses a processor to function.
Claim 12 is the parallel system claim of claim 2 and will be rejected under the same premise.
As per claim 13
Jeon and Yang teach all claim limitations previously rejected in claim 12’s 103 rejection. See claim 12’s 103 rejection.
Claim 13 is the parallel system claim of claim 3 and will be rejected under the same premise.
The Jeon/Yang method and its concepts necessarily use a computerized environment that uses a processor to function.
As per claim 14
Jeon and Yang teach all claim limitations previously rejected in claim 12’s 103 rejection. See claim 12’s 103 rejection.
Claim 14 is the parallel system claim of claim of 4 and will be rejected under the same premise.
The Jeon/Yang method and its concepts necessarily use a computerized environment that uses a processor to function.
As per claim 15
Claim 15 is the parallel system claim of claim 5 and will be rejected under the same premise.
Jeon and Yang teach all claim limitations previously rejected in claim 14’s 103 rejection. See claim 14’s 103 rejection.
The Jeon/Yang method and its concepts necessarily use a computerized environment that uses a processor to function.
Allowable Subject Matter
Claim 7-9 and 16-18 are 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.
Response to Arguments
Applicant’s arguments with respect to claims 1, 10 and 11 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 SHANE WRENSFORD CODRINGTON whose telephone number is (571)272-8130. The examiner can normally be reached 8:00am-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, Matthew Bella can be reached at (571) 272-7778. 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.
/SHANE WRENSFORD CODRINGTON/Examiner, Art Unit 2667
/MATTHEW C BELLA/Supervisory Patent Examiner, Art Unit 2667
Prosecution Insights