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
The Amendment filed June 23rd 2026 has been entered. Claims 1-30 are pending in the application. Claims 8, 18, and 28 are cancelled. Applicant’s amendments to the Claims 1, 11, and 21 have overcome the rejections previously set forth in the Non-Final Office Action mailed March 25th 2026. A further search has been performed to address the material amended in the aforementioned claims. Liu (NPL: Rapid Face Asset Acquisition with Recurrent Feature Alignment; hereinafter Liu 2022) was used for the amended limitations.
Response to Arguments
The Examiner appreciates the Applicant’s clarifications and discussion during the interview on June 10th 2026. The Examiner performed further search and found the Liu 2022 NPL reference. Although the Liu 2022 reference covers substantially the same subject matter as the Liu 2024 patent, the Examiner found that the Liu 2022 reference contains details not explicitly present in the Liu 2024 patent and vice versa (for example, whether certain “features” could be considered feature maps, or the description of residuals as maps as in claim 2 of the present application).
Applicant’s arguments with respect to claims 1, 11, and 21 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.
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.
Claims 1, 9, 10, 11, 19, 20, 21, 29, and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20210390789 A1; hereinafter Liu 2021) in view of Liu (NPL: Rapid Face Asset Acquisition with Recurrent Feature Alignment, hereinafter Liu 2022) and The Foundry Visionmongers Ltd (NPL: UV Mapping).
Regarding claim 1:
Liu 2021 teaches:
A method for generating a representation of a face (Liu 2021: The method can generate, based on the UV face position map, a 3D model of the face, Abstract), the method comprising:
obtaining a plurality of images of a face (Liu 2021: In some implementations, the image processing system 100 can perform face image augmentation and generate facial images with rotated faces from multiple images captured by the image sensor device [0046]);
extracting features for each image of the plurality of images (Liu 2021: For example, the compute components 110 can […] extract features and information (e.g., color information, texture information, semantic information, etc.) from the image data, etc. [0055])
Liu 2021 fails to teach:
extracting features for each image of the plurality of images to generate a plurality of feature maps; and
fusing the plurality of feature maps based on features of the plurality of feature maps along a common axis to generate an aligned feature map, wherein each feature map of the plurality of feature maps corresponds to a different image of the plurality of images.
Liu 2022 teaches:
A method for generating a representation of a face (Liu 2022: We present Recurrent Feature Alignment (ReFA), an end-to-end neural network for the very rapid creation of production-grade face assets from multi-view images, Pg. 1, par. 1), the method comprising:
obtaining a plurality of images of a face (Liu 2022: Starting from the multi-view images, we first reconstruct the geometry of the scan with neutral expression of the target subject using a multi-view stereo (MVS) algorithm, Pg. 4, Section 3.2: Data Preparation, par. 1);
extracting features for each image of the plurality of images to generate a plurality of feature maps (Liu 2022: we use a ResNet-like [He et al. 2016] backbone network to extract 2D features at 1/8 of the image resolution, Pg. 6, Section 4.2: Feature Extraction Networks; emphasis added; see Note 1A), wherein each feature map of the plurality of feature maps corresponds to a different image of the plurality of images (see Note 1B); and
fusing (Liu 2022: the flattened features at each view are fused by a chosen aggregation function, Pg. 7, col. 2, par. 2) the plurality of feature maps based on features of the plurality of feature maps along a common axis to generate an aligned feature map (Liu 2022: In Equation (2), our Visual Semantic Correlation (VSC) network (Section 4.3.1) matches the UV space feature and the image space feature, and produces a correlation feature map, Pg. 7, Compute Gradient, par. 2; emphasis added), wherein the common axis comprises a U, V axis (where ˜y(𝑡)𝑖, is the constructed 5D correlation tensor for 𝑖-th camera view, […] u is the 2D coordinates in the UV space; Pg. 7, col. 2, par. 2; see Note 1C), and wherein the aligned feature map represents features from at least two images of the plurality of images (Liu 2022: The correlation feature is a local representation of the alignment between the observed visual information and the semantic priors, see Note 1D; emphasis added).
Note 1A: Liu 2022 teaches that the feature extraction network extracts “2D features”, indicating that, as opposed to being a 1D vector or numerical data, the features may reasonably be understood to be 2D images or maps.
Note 1B: In Fig. 5 on Pg. 7, Liu showcases that an image space geometry feature is extracted from each image. One of ordinary skill in the art would interpret these image space geometry features to be feature maps, because they may be two-dimensional, as cited above: “we use a ResNet-like [He et al. 2016] backbone network to extract 2D features at 1/8 of the image resolution,” (Liu 2022, Pg. 6, Section 4.2: Feature Extraction Networks)
PNG
media_image1.png
290
856
media_image1.png
Greyscale
Cropped variant of Fig. 5 on Liu 2022 on Pg. 7.
Note that these Image space features are eventually used to generate the correlation feature (seen in the full Fig. 5 reproduced below). Because Liu 2022 depicts obtaining features from 3 images, the Examiner submits that Liu 2022 teaches that “the aligned feature map represents features from at least two images of the plurality of images” as claimed.
Note 1C: Liu 2022 teaches that the fused correlation feature map is represented by:
PNG
media_image2.png
93
159
media_image2.png
Greyscale
Fig. 5 of Liu 2022 on Pg. 7.
Liu 2022 further teaches that “u is the 2D coordinates in the UV space”, which is present as part of the definition of the Correlation feature above. The Foundry Visionmongers Ltd teaches: “A UV Map is a type of vertex map that stores vertical and horizontal positions on a 2D texture. The letters U (Horizontal) and V (Vertical) denote the axes of the 2D texture,” Pg. 1, par. 1.
It is shown that in Fig. 5, the Correlation feature above is created by a View aggregation function as shown below. Liu 2022 teaches that at least one feature, the “UV space feature” is fused based on an “Inner-product” with the image space features. See cropped variant of Fig. 5 of Liu 2022 below:
PNG
media_image3.png
380
571
media_image3.png
Greyscale
Cropped variant of Fig. 5 of Liu 2022 on Pg. 7.
Therefore, the Examiner submits that when the plurality of features are fused, they will be fused based on the Image space axis, as well as the UV space axis. It follows that the resulting “common axis” comprises at least the UV axis.
Note 1D: In Figure 5, Liu 2022 showcases that image features (i.e., Image feature view 1, Image feature view 2, Image feature view 3) from each view in image space is used to generate a correlation feature. Liu 2022 teaches: “The correlation feature is a local representation of the alignment between the observed visual information and the semantic priors.” (Pg. 7, Fig. 5).
PNG
media_image4.png
421
1457
media_image4.png
Greyscale
Fig. 3 of Liu 2022 on Pg. 7.
Before the effective filing date of the claimed invention, it would have been obvious to a person having ordinary skill in the art to combine the teachings of Liu 2022 with Liu 2021. Extracting features for each image of the plurality of images to generate a plurality of feature maps; and fusing the plurality of feature maps based on features of the plurality of feature maps along a common axis to generate an aligned feature map, as in Liu 2022, would benefit the Liu 2021 teachings by producing the face mesh more quickly: “ReFA is on par with the industrial pipelines in quality for producing accurate, complete, registered, and textured assets directly applicable to physically-based rendering, but produces the asset end-to-end, fully automatically at a significantly faster speed at 4.5 FPS, which is unprecedented among neural-based techniques.” (Liu 2022, Pg. 1, par. 1).
Furthermore, it would be obvious to combine the teachings of The Foundry Visionmongers Ltd. with the teachings of Liu 2021 and Liu 2022 because The Foundry Visionmongers Ltd. teaches the common understanding of the UV space that is well known in the art.
Regarding claim 9:
Liu 2021 in view of Liu 2022, and The Foundry Visionmongers Ltd teaches:
The method of claim 1 (as shown above), plurality of images of a face comprises views of the face from a plurality of angles around the face (Liu 2022: (a) Selected views of the captured images as input, Pg. 4, Fig. 2; see Note 9A).
Note 9A: Liu 2022 showcases in Fig. 2(a) on Pg. 4 that the input images comprise views of the face from a plurality of angles around the face.
Regarding claim 10:
Liu 2021 in view of Liu 2022, and The Foundry Visionmongers Ltd teaches:
The method of claim 1 (as shown above), wherein features for each image of the plurality of images are extracted using a set a machine learning based feature extractors (Liu 2022: subsequent sections are dedicated to the three main components of our system: (1) the feature extraction networks (Section 4.2) that extract features for the input images and a predefined UV-space feature map, Pg. 5, col. 2, par. 2).
Note 10A: Liu 2022 showcases in Fig. 3 that the feature extraction networks comprise encoders, which are known in the art to be components of encoder-decoder neural networks. In this case, the decoders are part of the “Learned reccurent face geometry optimizers” and the “Texture inference networks”, also shown in Fig. 3.
Regarding claim 11:
Claim 11 is substantially similar to claim 1, and is therefore rejected for similar reasons. Claim 11 contains the following notable differences:
Claim 11 claims an apparatus instead of a method. Liu 2021 teaches an apparatus: “an apparatus is provided for face image augmentation is provided” [0005].
Regarding claim 19:
Claim 19 is substantially similar to claim 9, and is therefore rejected for similar reasons. Claim 19 contains the following notable differences:
Claim 19 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 20:
Claim 20 is substantially similar to claim 10, and is therefore rejected for similar reasons. Claim 20 contains the following notable differences:
Claim 20 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 21:
Claim 21 is substantially similar to claim 1, and is therefore rejected for similar reasons. Claim 21 contains the following notable differences:
• Claim 21 claims a non-transitory computer-readable storage medium instead of a method. Liu 2021 teaches a non-transitory computer-readable medium: “a non-transitory computer-readable medium is provided for face image augmentation is provided. In some aspects, the non-transitory computer-readable medium can include instructions that, when executed by one or more processors, cause the one or more processors to …” [0006].
Regarding claim 29:
Claim 29 is substantially similar to claim 9, and is therefore rejected for similar reasons. Claim 29 contains the following notable differences:
Claim 29 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Regarding claim 30:
Claim 30 is substantially similar to claim 10, and is therefore rejected for similar reasons. Claim 30 contains the following notable differences:
• Claim 30 claims an apparatus instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches an apparatus.
Claims 2, 3, 4, 5, 6, 7, 12, 13, 14, 15, 16, 17, 22, 23, 24, 25, 26, and 27 are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20210390789 A1; hereinafter Liu 2021) in view of Liu (NPL: Rapid Face Asset Acquisition with Recurrent Feature Alignment, hereinafter Liu 2022), The Foundry Visionmongers Ltd (NPL: UV Mapping), and Liu 2024 (US 20240119671 A1).
Regarding claim 2:
Liu 2021 in view of Liu 2022 teaches:
The method of claim 1 (as shown above), further comprising:
obtaining a generic 3D morphable model (3DMM) of a face, the generic 3DMM including a plurality of vertices (Liu 2021: FIG. 5 illustrates an example 3D morphable head model, [0027]; see Note 2A);
projecting the generic 3DMM to two dimensions (Liu 2021: The 3DMM head shape can be projected onto an image plane with weak perspective projection, [0105]; see Note 2B) based on the common axis to generate a mean face position map (Liu 2022: The position map M is our representation of the face geometry. M comes with a UV mapping from a template mesh connectivity, and thus each pixel on M is mapped to a vertex or a surface point of a 3D mesh; see Note 2C; and Note 2D);
Note 2A: Liu 2021 teaches a 3DMM in Fig. 5, which comprises a face. Furthermore, the topology of the 3DMM is visible, wherein the vertices are marked by the intersections of the lines on the drawing. Therefore, the 3DMM includes a plurality of vertices.
Note 2B: A plane is two dimensional, so the 3DMM is projected to two dimensions.
Note 2C: As shown in Note 2B, the 3DMM is projected to a two dimensional plane. The specification of the present application recites: “The mean face position map may be 2D projection of a generic (e.g., mean, default) 3D mesh model of the object,” [0026]. Therefore, the Examiner submits that the 2D projection taught by Liu 2021 is analogous to the mean face position map.
Note 2D: Liu 2022 teaches a position map M in the UV space: “As shown in Figure 3, our end-to-end system takes multi-view images and a predefined template UV position map in a canonical space as input and produces 1) an updated position map” (Liu 2022; Pg. 5, Section 4: Method, par. 1). Because Liu 2022’s “position map M is our representation of the face geometry”, the Examiner submits that it would be obvious to utilize the 2D projection of the 3DMM taught in Liu 2021 as the position map in Liu 2022.
Liu 2021 in view of Liu 2022 and The Foundry Visionmongers Ltd. fails to explicitly teach:
determining one or more correspondences between features of the aligned feature map and vertices of the mean face position map to generate a first residual position map; and
combining the first residual position map with the mean face position map to generate an intermediate position map.
Liu 2024 teaches:
determining one or more correspondences between features of the aligned feature map and vertices of the mean face position map (Liu 2022: DIFF further updates the UV position map and the camera pose iteratively using an RNN-based neural optimizer from the feature correlation between the UV space and the image space [0080]; see Note 2E) to generate a first residual position map (Liu 2024: the residual of the position map δM [0101]; see Note 2E); and
combining the first residual position map (Liu 2024: To predict the update tuple, according to embodiments, a 2D feature map is constructed containing the signals where δM(t) and [δR(t), δt(t)] should orient, [0134]; see Note 2F) with the mean face position map (Liu 2024: Specifically, let M̂(t) = R(t) M(t) + t(t) be the transformed position map at t-th step, [0135]; see Note 2G) to generate an intermediate position map (see Note 2G).
Note 2E: Liu 2024 teaches that “deep iterative face fitting (DIFF), a non-parametric approach based on feature correlation” [0080] will be used to update the UV position map (as in [0080] cited above) based on a feature correlation between the UV space and image space. Liu 2024 further teaches: “FIG. 4 illustrates an overview of an exemplary pipeline for a given input image when processed by the disclosed DIFF method and system,” [0085]. In Fig. 4, Liu 2024 showcases that VSC receives input from the “UV-space feature map G” and “M(0)”, where M is the mean face position map (as shown in Note 2F below). Because DIFF generates correlation data between the mean face position map and the UV-space feature map, Liu 2024 teaches determining one or more correspondences between features of the aligned feature map and vertices of the mean face position map.
Liu 2024 teaches: “the resulting feature vector y(t) may be used as the output of the visual semantic correlation network. Liu 2024 further teaches: “Given the hidden state h(t), the geometry decoding network outputs […] the position correction map δM(t),” [0101]. The hidden state is calculated by the following equation (Liu 2024, [0131]):
PNG
media_image5.png
24
378
media_image5.png
Greyscale
That is, the hidden state at step t is based on the feature vector from the correlation, and the previous hidden state at step t-1.
Because Liu teaches that the position correction map is analogous to a residual of the position map: “the residual of the position map δM,” [0101], and the position correction map is generated based on the hidden state obtained from the feature vector representing the correlation data, Liu teaches determining one or more correspondences between features of the aligned feature map and vertices of the mean face position map to generate a first residual position map.
Note 2F: Liu 2024 teaches: “the position map M may be updated as well as the head pose [R, t] separately, given the correlation tensor between the two misaligned feature maps of interest, namely the UV feature map g and the image space feature f,” [0131]. As cited above, Liu further teaches: “according to embodiments, a 2D feature map is constructed containing the signals where δM(t) and [δR(t), δt(t)] should orient,” [0134]. Note that δM(t) is the residual feature map cited in (Liu 2024, [0101]) above. Because the residual position map is used to calculate a feature map that is then used to update the position map (further details on the updating in Note 2F), Liu 2024 teaches that the first residual position map may be combined with the mean face position map to generate an updated position map.
Note 2G: Liu 2024 teaches that the position map M is updated according to this equation: “In the following paragraphs, there is described in detail how the disclosed optimizer initializes, updates, and finalizes the corrections in order to recover M…” [0133]. Liu 2024 teaches: “A UV-space position map that represents the geometry may be first initialized to be the mean face shape, according to embodiments,” [0129]. That is, prior to updating, the position map M is initialized to be a mean face position map. As cited above, the position map is updated to “M̂(t)”, which is analogous to the intermediate position map.
Before the effective filing date of the claimed invention, it would have been obvious to a person having ordinary skill in the art to combine the teachings of Liu 2024 with Liu 2021 in view of Liu 2022 and The Foundry Visionmongers Ltd.. Determining one or more correspondences between features of the aligned feature map and vertices of the mean face position map to generate a first residual position map; and combining the first residual position map with the mean face position map to generate an intermediate position map., as in Liu 2024, would benefit the Liu 2021, Liu 2022, and The Foundry Visionmongers Ltd. teachings by ensuring that features are matched as accurately as possible while avoiding overfitting to specific features or geometry. (Liu 2024: Such results may be achieved with less than 200 training subjects, which demonstrates the generalizability of the disclosed model across identity, thanks to iterative correlation design that learns the local semantic features, [0113])
Regarding claim 3:
Liu 2021 in view of Liu 2022, The Foundry Visionmongers Ltd., and Liu 2024 teaches:
The method of claim 2 (as shown above), wherein
determining the one or more correspondences comprises determining one or more displacement values (Liu 2024: The first is the use of a UV-space position map for representing geometry, where each pixel is mapped to the position of a surface vertex. Such representation […] organically aligns the geometry and texture space for the inference of high-frequency displacement maps ... [0062]; Liu 2024: The recurrent optimization is centered around a per-pixel visual semantic correlation (VSC) that serves to iteratively refine the face geometry, [0062]; see Note 3A) between features of the aligned feature map and vertices of the mean face position map (see Note 3B), and
wherein the first residual position map includes the one or more displacement values (see Note 3A).
Note 3A: In Fig. 4 of Liu 2024, it is shown that VSC accepts the position map M as input. Given that the position map comprises vertex positions, updating the position map inherently requires displacing the vertices of the position map.
Furthermore, as shown above in Note 2F, the position map M is updated to “M̂(t)”. Therefore, any displacement of the vertices will be encoded in M̂(t). Because Liu 2024 “iteratively comput[es] […] a feature map via visual semantic correlation between the UV space and the image space and regress geometry updates,” [0021], it is reasonable to conclude that the displacement encoded in the position map at step t will remain when the system completes further iterations.
Note 3B: As shown in Note 2D, the one or more correspondences are determined based on features of the aligned feature map and vertices of the mean face position map.
Regarding claim 4:
Liu 2021 in view of Liu 2022, The Foundry Visionmongers Ltd., and Liu 2024 teaches:
The method of claim 2 (as shown above), wherein the one or more correspondences are also determined based on a label map labeling portions of the mean face position map (see Note 4A).
Note 4A: Fig. 23 of Liu 2024 showcases that landmarks may be assigned to the position map. Specifically, Liu 2024 teaches that “Given valid UV mappings, the position map representation is amenable to conversions to various representations, as shown in each column. Specifically, the disclosed position map may be converted to different mesh topologies seamlessly as long as a solid UV mapping is provided,” [0170]. The sixth column from the left showcases landmarks, which is analogous to a label map that labels portions of the mean face position map.
It would be obvious to one of ordinary skill in the art to also determine correspondences based on the landmarks, because Liu teaches: “Landmarks and region maps representations can be extracted from the potential maps, which are suitable for many mobile applications,” [0170].
Regarding claim 5:
Liu 2021 in view of Liu 2022, The Foundry Visionmongers Ltd., and Liu 2024 teaches:
The method of claim 2 (as shown above), further comprising:
determining one or more correspondences between features of the aligned feature map and vertices of the mean face position map (Liu 2024: DIFF further updates the UV position map and the camera pose iteratively using an RNN-based neural optimizer from the feature correlation between the UV space and the image space [0080]; see Note 2D and Note 5A) to generate a second residual position map (Liu 2024: the residual of the position map δM [0101]; see Note 2D and Note 5A); and
combining the second residual position map (Liu 2024: To predict the update tuple, according to embodiments, a 2D feature map is constructed containing the signals where δM(t) and [δR(t), δt(t)] should orient, [0134]; see Note 2E and Note 5A) with the mean face position map (Liu 2024: Specifically, let M̂(t) = R(t) M(t) + t(t) be the transformed position map at t-th step, [0135]; see Note 2F and Note 5A) to generate a fine position map (see Note 2F and Note 5A).
Note 5A: As noted in Note 3A, Liu 2024 “iteratively comput[es] […] a feature map via visual semantic correlation between the UV space and the image space and regress geometry updates,” [0021]. That is, Liu may determine correlations or “correspondences” an indefinite amount of times (Liu 2024, Fig. 4, “Learned recurrent face geometry optimizers”), before outputting “refined geometry” [0062].
Regarding claim 6:
Liu 2021 in view of Liu 2022, The Foundry Visionmongers Ltd., and Liu 2024 teaches:
The method of claim 5 (as shown above), further comprising reprojecting the fine position map to three dimensions to obtain a fine face mesh (Liu 2021: In some cases, generating the 3D model of the face can include projecting the UV face position map to 3D space [0011]).
Regarding claim 7:
Liu 2021 in view of Liu 2022, The Foundry Visionmongers Ltd., and Liu 2024 teaches:
The method of claim 6 (as shown above), further comprising:
aligning textures of the plurality of images based on the fusing of the plurality of feature maps to generate a texture map (Liu 2022: The refined geometry then provides pixel-aligned signals to a texture inference network that accurately infers albedo, specular and displacement map in the same UV space; Pg. 3, par. 1); and
applying the texture map to the fine face mesh to obtain a representation of the face (Liu 2022, Pg. 9, Fig. 6; see Note 7A).
Note 7A: Liu 2022 showcases “Images rendered from our reconstructed face assets.” (Pg. 9, Fig. 6). The faces are rendered with the applied textures/inferred maps.
Regarding claim 12:
Claim 12 is substantially similar to claim 2, and is therefore rejected for similar reasons. Claim 12 contains the following notable differences:
Claim 12 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 13:
Claim 13 is substantially similar to claim 3, and is therefore rejected for similar reasons. Claim 13 contains the following notable differences:
Claim 13 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 14:
Claim 14 is substantially similar to claim 4, and is therefore rejected for similar reasons. Claim 14 contains the following notable differences:
Claim 14 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 15:
Claim 15 is substantially similar to claim 5, and is therefore rejected for similar reasons. Claim 15 contains the following notable differences:
Claim 15 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 16:
Claim 16 is substantially similar to claim 6, and is therefore rejected for similar reasons. Claim 16 contains the following notable differences:
Claim 16 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 17:
Claim 17 is substantially similar to claim 2, and is therefore rejected for similar reasons. Claim 17 contains the following notable differences:
Claim 17 claims an apparatus instead of a method. In the rejection of claim 11, it was shown that Liu 2021 teaches an apparatus.
Regarding claim 22:
Claim 22 is substantially similar to claim 2, and is therefore rejected for similar reasons. Claim 22 contains the following notable differences:
Claim 22 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Regarding claim 23:
Claim 23 is substantially similar to claim 3, and is therefore rejected for similar reasons. Claim 23 contains the following notable differences:
Claim 23 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Regarding claim 24:
Claim 24 is substantially similar to claim 4, and is therefore rejected for similar reasons. Claim 24 contains the following notable differences:
Claim 24 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Regarding claim 25:
Claim 25 is substantially similar to claim 5, and is therefore rejected for similar reasons. Claim 25 contains the following notable differences:
Claim 25 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Regarding claim 26:
Claim 26 is substantially similar to claim 6, and is therefore rejected for similar reasons. Claim 26 contains the following notable differences:
Claim 26 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Regarding claim 27:
Claim 27 is substantially similar to claim 2, and is therefore rejected for similar reasons. Claim 27 contains the following notable differences:
Claim 27 claims a non-transitory computer-readable storage medium instead of a method. In the rejection of claim 21, it was shown that Liu 2021 teaches a non-transitory computer-readable medium.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Yang (US 11417053 B1) showcases a similar refinement method for the position map in Fig. 2.
Wu (NPL: MVF-Net: Multi-View 3D Face Morphable Model Regression) teaches reconstruction of a face from multiple views utilizing features obtained from each input image.
Yao (NPL: MVSNet: Depth Inference for Unstructured Multi-view Stereo) was previously cited to teach “extracting features for each image of the plurality of images to generate a plurality of feature maps that correspond to the plurality of images”.
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 VINCENT ALEXANDER PROVIDENCE whose telephone number is (571)270-5765. The examiner can normally be reached Monday-Thursday 8:30-5:00.
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, King Poon can be reached at (571)270-0728. 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.
/VINCENT ALEXANDER PROVIDENCE/Examiner, Art Unit 2617 /KING Y POON/Supervisory Patent Examiner, Art Unit 2617
Prosecution Insights