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 .
This action is in response to the Amendment filed on 8/12/2026.
Claims 1-5, 7-15, 17-20 are pending. Claims 1, 7, 9, 11, 17 have been amended. Claim 6, 16 were cancelled.
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) 1-6, 8-16, 18-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu et al (US 20240119671 A1, hereinafter Liu) in view of Yu et al. (US 10991145 B2, hereinafter Yu) and further in view of Jing et al., "Uncertainty Guided Adaptive Warping for Robust and Efficient Stereo Matching," arXiv:2307.14071, June 2023 (hereinafter Jing).
Regarding Claim 1, Liu teaches A method for controlling an artificial intelligence (AI) device, the method comprising: (Liu, Abstract, "Methods, mediums, and systems for constructing a 3D face model, including Deep Iterative Face Fitting (DIFF) and Recurrent Feature Alignment (ReFA)"), receiving, by a processor, an input two-dimensional (2D) image; (Liu, Abstract, "receives input image(s)—as a template and UV texture mapping pointing the 3D mesh to a 2D UV space; receives input image(s)"), encoding, by the processor, the 2D image to generate an image feature map; (Liu, Paragraph [0089], "the feature G is processed through a convolutional neural network and get the resulting UV feature map"; it is noted since the encoding the input image through a neural network backbone to extract feature representations at reduced spatial resolution, which reads on generating an image feature map), [[obtaining, by the processor, UV positional encoding information based on a three-dimensional (3D) face model, the UV positional encoding information including a UV feature map ]], generating, by the processor, a correlation volume based on the image feature map and the UV feature map; (Liu, Paragraph [0135], "the 2D feature map computes similarity between the multi-view geometry features f and the UV features g by constructing a correlation volume. Specifically, let {circumflex over (M)}(t)=R(t)M(t)+t(t) be the transformed position map"; it is noted the correlation volume is explicitly built from image features g and template features derived from the UV feature map G), generating, by the processor, flow map information and [[uncertainty information]] based on the correlation volume and the UV positional encoding information; (Liu, Paragraph [0061], "the disclosed network may extract per-pixel features in both the multi-view image space and the UV space and a recurrent module then iteratively optimizes the geometry by projecting the image-space features to the UV space and comparing them with a reference UV-space feature"; Liu, Paragraph [0090], "this network iteratively optimizes the face geometry and outputs the pose R, t and the position map M(t)"; "One of the central modules of the disclosed recurrent face geometry optimizer is a recurrent neural network (RNN) composed of Gated Recurrent Units (GRU)"), and generating, by the processor, probabilistic 2D alignment information based on the flow map information and [[uncertainty information]] and outputting the probabilistic 2D alignment information, (Liu, Paragraph [0019], "predicting the probabilistic distributions of each individual vertex in a coarse-to-fine manner"; Liu, Paragraph [0097], "The visual-semantic similarities may be calculated as the inner-product between the geometry feature f and the UV feature g, in accordance with an embodiment, called Visual-Semantic Correlation (VSC) herein" <read on probabilistic 2D alignment information>), wherein the generating the flow map information and the [[uncertainty information]] comprises: iteratively refining, via a recurrent update block based on a neural network, a UV-to-image flow estimate and [[an uncertainty estimate]] based on the recurrent update block receiving, for each iteration, information from the correlation volume, a context map, a previous hidden state, a previous flow estimate, and [[a previous uncertainty estimate]] to generate a refined flow estimate, [[a refined uncertainty estimate]], and an updated hidden state; (Liu, Paragraph [0090], "At the t-th step, in accordance with some embodiments herein, the network update may be written as: y(t)←VSC(f, g, R(t-1), t(t-1), M(t-1)), (3) h(t)←GRU(y(t), h(t-1)), (4) Δ(t)←Decoder(h(t)), (5)"; Liu, Paragraph [0164], "GRU layers take the input of the current step and the hidden layer from the previous step as inputs, and outputs the updated hidden layer" <read on receiving a previous hidden state and outputting an updated hidden state>), and generating the flow map information and the [[uncertainty information]] based on the refined flow estimate, [[the refined uncertainty estimate]], and the updated hidden state (Liu, Paragraph [0091], "the geometry decoding network (Section 4.3.2) is used to process the hidden vector h(t) and compute the geometry update tuple Δ(t)" <read on generating the flow map information based on the refined flow estimate and the updated hidden state>).
But Liu does not explicitly disclose obtaining, by the processor, UV positional encoding information based on a three-dimensional (3D) face model, the UV positional encoding information including a UV feature map as a named, standalone encoding stage; … uncertainty information; … an uncertainty estimate, a previous uncertainty estimate, and a refined uncertainty estimate as inputs and outputs of the recurrent update block at each iteration.
However, Yu teaches obtaining, by the processor, UV positional encoding information based on a three-dimensional (3D) face model, the UV positional encoding information including a UV feature map (Yu, Column 2, Line 67 – Column 3, Line 5, "In the first stage, we apply the UV representation for both 3D point cloud and the texture, termed UVpos and UVtex respectively. The UV representation provides the dense shape information and also builds the one-to-one correspondence from point cloud to the texture"; Yu, Column 6, Line 1-5, "A UV position map UV, is defined on the UV space, each entry is the corresponding three dimensional coordinate (x, y, Z). A UV texture map UV, is also defined on the UV space, each entry is the corresponding coordinate's RGB color" <read on UV feature map>).
Yu and Liu are analogous since both deal with 3D face model processing from monocular 2D images, UV-space mapping of 3D geometry, and neural network-based face reconstruction pipelines. Liu provided a method of recurrent feature alignment using a UV feature map extracted from the 3D template mesh. Yu provided a method of constructing an explicit UV positional map (UVpos) from the 3D point cloud of the face model to encode geometric structure in UV space and to map 3D geometry into UV space for one-to-one 3D-to-2D correspondence. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the explicit UV positional encoding (UVpos) approach taught by Yu into the invention of Liu such that the UV positional encoding information is derived from the 3D face model as a structured UV feature map. The motivation is to provide dense shape information and a one-to-one correspondence from the 3D point cloud to the image domain.
However, Jing teaches uncertainty information (Jing, Section 3.2, page 3, "Given the cost volume Vn, the uncertainty map Un is estimated and fed to the UGAC of the next iteration"; Jing, Section 3.2, page 4, "we introduce a variance-based uncertainty estimation to guide the offset o(p, k), and further balance the disparity distributions of different datasets"), an uncertainty estimate, a previous uncertainty estimate, and a refined uncertainty estimate as inputs and outputs of the recurrent update block at each iteration (Jing, Section 3.1, page 3, "the multi-scale features pass through 3 cascaded stages of the proposed adaptive recurrent module (ARM), which is composed of an uncertainty guided adaptive correlation (UGAC) module and a gate recurrent unit (GRU)" <read on a recurrent update block based on a neural network>; "In the ARM, the cost volume is calculated via UGAC and then input into GRU, for iteratively refining the disparity prediction results" <read on iteratively refining ... a UV-to-image flow estimate>; Jing, Section 3.2, page 3, " For the n-th iteration of ARM, the right features {FRs} are first warped via the content-aware warp-ing layer, considering the prediction disparity dn−1 and the uncertainty map Un−1 at the (n − 1)-th iteration" <read on the recurrent update block receiving, for each iteration, ... a previous flow estimate, and a previous uncertainty estimate>; "Given the cost volume Vn, the uncertainty map Un is estimated and fed to the UGAC of the next iteration" <read on generate ... a refined uncertainty estimate>; "Note that Vn is used as the input of GRU" <read on receiving ... information from the correlation volume>).
Jing and Liu are analogous since both address iterative dense correspondence estimation between two feature maps using a recurrent update block based on a Gated Recurrent Unit (GRU) that consumes a correlation/cost volume and iteratively refines a per-pixel offset field over multiple iterations. Liu provided a GRU-based recurrent feature alignment framework (ReFA) that iteratively refines pose and a UV-space position map using a correlation feature map, a context map, and a carried hidden state, but does not explicitly feed a previous uncertainty estimate back into the recurrent block as a distinct per-iteration input. Jing provided an Adaptive Recurrent Module (ARM) composed of an Uncertainty Guided Adaptive Correlation (UGAC) module and a GRU, in which a variance-based uncertainty map produced at the (n−1)-th iteration is explicitly fed back as an input to the n-th iteration of the recurrent update block, so that the block simultaneously refines the correspondence estimate and the uncertainty estimate at each iteration. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the per-iteration uncertainty feedback mechanism taught by Jing into the GRU-based recurrent face geometry optimizer of Liu such that the recurrent update block receives, at each iteration, the correlation volume, a context map, a previous hidden state, a previous flow estimate, and a previous uncertainty estimate, and outputs a refined flow estimate, a refined uncertainty estimate, and an updated hidden state. The motivation is to improve robustness in ill-posed regions such as self-occluded, texture-less, and ambiguous areas that inherently arise when estimating per-vertex 2D positions of a 3D face under pose variation and expression changes, as discussed by Jing at Section 3.2, page 4.
Regarding Claim 2, the combination of Liu, Yu and Jing teaches the invention in Claim 1.
The combination further teaches the probabilistic 2D alignment information includes information about points on the 3D face model, a predicted 2D location, and a corresponding uncertainty value: (Liu, Paragraph [0019], "predicting the probabilistic distributions of each individual vertex in a coarse-to-fine manner"; [0021], “a 3D mesh that represents an item and a UV texture mapping that maps each point on the 3D mesh to a 2D UV space, the item being a median face, vertex positions or surface points;”; “a coordinate of a corresponding point in the canonical space of the 3D mesh”).
Jing teaches a corresponding uncertainty value (Jing, Section 3.2, page 4, Equation (4), "Un = 1 − σ(P (¯Vn − Vn)^2)"; "we introduce a variance-based uncertainty estimation to guide the offset o(p, k)").
As explained in the rejection of Claim 1, the obviousness for combining the per-iteration uncertainty feedback of Jing into Liu is provided above.
Regarding Claim 3 the combination of Liu, Yu and Jing teaches the invention in Claim 1.
The combination further teaches generating 3D reconstruction information based on the probabilistic 2D alignment information for animating a 3D face based on the input 2D image: (Liu, Paragraph [0058], "this disclosure presents a Deep Iterative Face Fitting (DIFF), an end-to-end neural network for face asset creation. For example, in embodiments, from a single image, a method is disclosed that is designed to reconstruct a complete face model, including pore-level face geometries along with detailed albedo and specular maps in 4K resolution"; Liu, Paragraph [0064], "fully automatically create a 3D face avatar that is production-ready. The disclosed model outperforms previous neural-based approaches in both accuracy and speed" <read on generating 3D reconstruction information for animating a 3D face based on the input 2D image>).
Regarding Claim 4 the combination of Liu, Yu and Jing teaches the invention in Claim 3.
The combination further teaches displaying a 3D facial animation with animated movements based on the 3D reconstruction information (Liu, Paragraph [0054], "face modeling significantly boosts the development of many applications, including AR/VR, teleconference, animation in gaming, and social media apps that require a personalized avatar"; [0057], "produce complete face assets for industrial rendering, animation pipeline for realistic rendering" [0069], "display DS may be an output device for outputting a pose or updated 3D template map, as described herein"; it is noted the updated 3D mesh output enables rendering of a 3D facial animation with animated movements).
Regarding Claim 5 the combination of Liu, Yu and Jing teaches the invention in Claim 3.
The combination further teaches the generating the 3D reconstruction information includes optimizing an energy function that includes one or more of an alignment energy component, a prior energy component, a temporal smoothness component, a 3D neural geometry component, and a deformation energy component: (Liu, Paragraph [0083], "image space optical flow from neutral to expression scan may be added as dense consistency constraint in the expression non-rigid Laplacian deformation to retain texture" <read on deformation energy component>; [0080], " deep iterative face fitting (DIFF), a non-parametric approach based on feature correlation and neural optimizers. Given a single image, DIFF infers highly accurate geometry and textures to generate a full set of production-ready assets. The disclosed network adopts a UV position map as the optimization domain, enabling pixel-level freedom to model non-linear deformations details“ <read on optimizing an energy function>; it is noted the DIFF system uses iterative optimization combining multiple energy/loss terms including alignment and deformation constraints).
Regarding Claim 8 the combination of Liu, Yu and Jing teaches the invention in Claim 1.
The combination further teaches the correlation volume is a 4D correlation volume in a form of a tensor: (Liu, Paragraph [0135], "the 2D feature map computes similarity between the multi-view geometry features f and the UV features g by constructing a correlation volume. Specifically, let {circumflex over (M)}(t)=R(t)M(t)+t(t) be the transformed position map at t-th step, this disclosure first enumerates a 3D regular grid of size (2r+1)×(2r+1)×(2r+1) around M(t)(u) for each pixel u in the UV space" <read on a 4D correlation volume in a form of a tensor — the inner-product similarity computed between the 2D UV feature map g and the 2D image feature map f across a 3D search grid yields a multi-dimensional tensor output for each UV pixel u>).
Regarding Claim 9 the combination of Liu, Yu and Jing teaches the invention in Claim 1.
The combination further teaches wherein the [[ uncertainty ]] information includes 2D Gaussian information indicating a measure of confidence for a prediction of a 2D position of a vertex in the 3D face model (Liu, Paragraph [0019], "predicting the probabilistic distributions of each individual vertex in a coarse-to-fine manner"; [0089], “from the template mesh and its UV mapping, the UV-space feature map G∈ Wt×Ht×36 may be assembled by concatenating the following features for each pixel u: (1) the 2D coordinate of u itself, normalized to [−1,1]2; (2) the corresponding 3D coordinate of the pixel u in the template face mesh"; it is noted the probabilistic distributions per vertex indicate confidence measures for predicted 2D positions of vertices in the 3D face model).
Yu further teaches 2D Gaussian information indicating a measure of confidence for predicted vertex positions (Yu, Column 3, Line 1-5, "The UV representation provides the dense shape information and also builds the one-to-one correspondence from point cloud to the texture"; Column 5, Line 6-10, “directly generate, by a 3D regression network, a space position map for a 3D shape and a camera perspective matrix from a single input image of a face”).
Liu and Yu are analogous since both address 3D face model-based reconstruction from single 2D input images, mapping 3D geometry into UV-parameterized 2D representations. Liu provided an iterative GRU-based alignment pipeline that outputs a UV-space position map per vertex but does not explicitly characterize the uncertainty of predicted vertex positions as 2D Gaussian distributions with a confidence measure. Yu provided a UV positional representation (UVpos) derived from a 3D point cloud that encodes dense per-point geometric information in UV space, establishing a structured 2D representation of 3D face model geometry. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the per-vertex probabilistic uncertainty characterization of Yu into the vertex prediction output of Liu such that the uncertainty information for each predicted 2D position of a vertex in the 3D face model is expressed as 2D Gaussian information indicating a measure of confidence. The motivation is to provide a principled probabilistic measure of prediction reliability per vertex leveraging the one-to-one UV-space so as to enable downstream processes to weight or filter predictions based on confidence.
Jing teaches 2D Gaussian information indicating a measure of confidence (Jing, Section 3.2, page 4, Equation (4), "Un = 1 − σ(P (¯Vn − Vn)^2)"; "we introduce a variance-based uncertainty estimation" <read on 2D Gaussian information indicating a measure of confidence>; Jing, Section 3.2, page 4, "σ(·) is the sigmoid function").
As explained in the rejection of Claim 1, the obviousness for combining the variance-based per-iteration uncertainty estimation of Jing into Liu is provided above.
Regarding Claim 10, the combination of Liu, Yu and Jing teaches the invention in Claim 1.
The combination further teaches the UV feature map is a representation of the 3D face model in UV space, the UV space being a 2D coordinate system for mapping points on the 3D face model to points on the input 2D image: (Liu, Abstract, "UV texture mapping pointing the 3D mesh to a 2D UV space"; "UV feature map G∈ℝ^(Wt/8×Ht/8×C)"; "3D mesh/UV-space position map is output, each pixel in the UV-space map storing a coordinate of a corresponding point in a canonical space of the 3D mesh"). UV space is explicitly described as the 2D coordinate system mapping 3D mesh points to 2D space; Paragraph [0089], " from the template mesh and its UV mapping, the UV-space feature map Ge R "#736 may be assembled by concatenating the following features for each pixel u: (1) the 2D coordinate of u itself, normalized to [-1,1]% (2) the corresponding 3D coordinate of the pixel u in the template face mesh")
Yu further reinforces: (Yu, Column 2, Line 67-Column 3, Line 1-5, "we apply the UV representation for both 3D point cloud and the texture, termed UVpos and UVtex respectively. The UV representation provides the dense shape information and also builds the one-to-one correspondence from point cloud to the texture").
As explained in rejection of claim 1, the obviousness for combining Yu into Liu is provided above.
Regarding Claim 11, it recites limitations similar in scope to the limitations of claim 1, but in a AI device.. As shown in the rejection, the combination of Liu, Yu and Jing disclose the limitations of claims 1. Additionally, Liu discloses an system that maps to Fig. 1 and Paragraph [0090]-[0091] (Liu, Fig. 1 Paragraph [0061], [0066]-[0068], FIG. 1 shows an exemplary computer system CS… that includes a bus BS or other communication mechanism for communicating information and at least one processor PRO ( or multiple processors PRO-1 and PRO-2) coupled with bus BS for processing information… Computer system CS also includes a main memory MM, such as a random access memory (RAM) or other dynamic storage device, coupled to bus BS for storing information and instructions to be executed by processor… The disclosed embodiments may also enable massive database capturing for deep learning purposes”). Thus, Claim 11 is met by Liu according to the mapping presented in the rejection of claims 1, given the method corresponds to the AI/deep learning device.
Regarding Claim 12, it recites limitations similar in scope to the limitations of Claim 2 and therefore is rejected under the same rationale.
Regarding Claim 13, it recites limitations similar in scope to the limitations of Claim 3 and therefore is rejected under the same rationale.
Regarding Claim 14, it recites limitations similar in scope to the limitations of Claim 4 and therefore is rejected under the same rationale.
Regarding Claim 15, it recites limitations similar in scope to the limitations of Claim 5 and therefore is rejected under the same rationale.
Regarding Claim 18, it recites limitations similar in scope to the limitations of Claim 8 and therefore is rejected under the same rationale.
Regarding Claim 19, it recites limitations similar in scope to the limitations of Claim 9 and therefore is rejected under the same rationale.
Regarding Claim 20, it recites limitations similar in scope to the limitations of Claim 10 and therefore is rejected under the same rationale.
Claim(s) 7, 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu et al (US 20240119671 A1, hereinafter Liu) in view of Yu et al. (US 10991145 B2, hereinafter Yu) and further in view of Jing et al., "Uncertainty Guided Adaptive Warping for Robust and Efficient Stereo Matching," arXiv:2307.14071, June 2023 (hereinafter Jing) as applied to Claim 1, 11 above respectively and further in view of Elbaz (US 20230044644 A1).
Regarding Claim 7, the combination of Liu, Yu and Jing teaches the invention in Claim 6.
The combination does not explicitly disclose but Elbaz teaches neural network is trained based on a dual loss function (Elbaz, Paragraph [0081], "Two different loss functions may be used for back-propagation to train the texture map autoencoder, a weighted L2 (square-integrable) loss function 132, and a Kullback—Leibler divergence (KL) loss function 134") that incorporates Gaussian negative loglikelihood (GNLL) (Elbaz, Paragraph [0082], "The KL loss function 134 may be used for training. This function enforces a unit Gaussian prior N (0, 1) with zero mean on the distribution of latent space vectors, meaning that this forces the latent space to be a multivariate, regular Gaussian distribution. Note that each value in the latent space is represented as a mean and variance"; it is noted the KL divergence against a Gaussian prior N(0,1) is mathematically equivalent to a form of Gaussian negative log-likelihood (GNLL) when optimizing a variational objective over a Gaussian distribution).
Elbaz and Liu are analogous since all of them deal with neural network-based 3D face modeling and reconstruction from 2D images, and all use multi-component training loss functions to optimize different aspects of the network's output. Liu provided a GRU-based recurrent alignment network. Elbaz provided a dual loss training paradigm combining a reconstruction-quality loss with a Gaussian-distribution-enforcing probabilistic loss. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the dual loss training approach of Elbaz into the modified invention of Liu such that system will be able to simultaneously supervise both the accuracy of the flow estimate and the calibration of the predicted uncertainty distribution. The motivation is to produce a well-calibrated Gaussian uncertainty output from the recurrent network, as the use of a Gaussian-enforcing loss (such as KL divergence or negative log-likelihood of a Gaussian) alongside a regression loss is a well-known dual-loss training strategy for networks that predict both a value estimate and its associated uncertainty.
Regarding Claim 17, it recites limitations similar in scope to the limitations of Claim 7 and therefore is rejected under the same rationale.
Response to Arguments
Applicant’s arguments with respect to claim 1, 11, 17, filed on 8/12/2026., with respect to rejection under 35 USC § 103 have been considered but are moot in view of the new ground(s) of rejection. it has now been taught by the combination of Liu, Yu and Jing.
In regard to Claims 2-5, 6-10, 12-15, 18-20, they directly/indirectly depends on independent Claim 1, 11, 17 respectively. Applicant does not argue anything other than the independent claim 1, 11, 17. The limitations in those claims in conjunction with combination previously established as explained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 20220398747 A1 VOLUMETRIC SAMPLING WITH CORRELATIVE CHARACTERIZATION FOR DENSE ESTIMATION
US 20210365786 A1 NEURAL NETWORK IMAGE ANALYSIS
US 20210287430 A1 SELF-SUPERVISED SINGLE-VIEW 3D RECONSTRUCTION VIA SEMANTIC CONSISTENCY
US 20210150203 A1 PARAMETRIC TOP-VIEW REPRESENTATION OF COMPLEX ROAD SCENES
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 YUJANG TSWEI whose telephone number is (571)272-6669. The examiner can normally be reached 8:30am-5:30pm EST.
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, Kent Chang can be reached on (571)272-7667. 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.
/YuJang Tswei/Primary Examiner, Art Unit 2614
Prosecution Insights