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
Claims 1-20 are pending. Claims 1, 8, and 15 are amended.
Response to Arguments
Applicant's arguments filed 11/13/2025 have been fully considered but they are not persuasive. Applicant argues,
Demyanov does not teach to use one or more neural networks to generate one or more images of one or more objects based, at least in part, on one or more reference frames of a reference video and at least two source images of the one or more objects, wherein an individual image of the one or more generated images is generated based, at least in part, on modifying different features of the one or more objects from different images of the at least two source images to be combined in the individual generated image to imitate features of an individual reference frame of the one or more reference frames of the reference video. In Demyanov only one of the source images is modified to generate each modified source image. To determine how to modify a single source image, Demyanov uses motion estimation differences between an identified driving image and an identified expression image (which the Office Action incorrectly maps to the source images of Applicant's claim). However, different features from the identified driving image and the identified expression image are not modified to be combined in the individual generated image to imitate features of an individual reference frame (which the Office Action incorrectly maps to the source image in Demyanov). Only a single source image is modified to generate one of the modified images in Demyanov. The identified driving image and the identified expression image are used to determine motion estimation differences, but features of those images to are not modified to be combined in an individual generated image. Demyanov col 21, lines 56-60 ("The image transformation neural network modifies a corresponding source image in the source image sequence by computing motion estimation differences between the identified driving image and the identified expression image.").
For at least the reasons presented above claims 1-7 are not anticipated by Demyanov. Similar reasons apply to claims 8-20. Accordingly, withdrawal of the
rejection and allowance of the application are respectfully requested.
Examiner does not agree with Applicant’s arguments and conclusions drawn therefrom. First and foremost according to fig. 9 of Demyanov, source image sequence in step 902 (mapped to reference frames and reference video claim 1) and driving image data of step 904 and expression image data of step 906 (mapped to two source images of claim 1) generate a modified source image sequence in step 908 (mapped to generated images, and individual images of the generated image of claim 1). Therefore, based on fig. 9 it is clear that two source images (driving image and expression image) and source image sequence is used to generate modified source image sequence. Furthermore, Demyanov also reasonably discloses that features of driving image and expression image are also modified to be combined in an individual generated image. See the following disclosures –
In Col. 21, lines 25-27, Demyanov discloses - “At operation 904, the facial animation system 700 identifies driving image sequence data to modify face image feature data in the source image sequence.” – confirming that driving image data indeed modifies face image features in the source image sequence.
In Col. 21, lines 25-27, Demyanov discloses – “At operation 906, the facial animation system 700 identifies an expression dataset to modify face image feature data in the source image sequence” – confirming that driving image data indeed modifies face image features in the source image sequence.
In Col. 21, lines 48-60, Demyanov discloses – “The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence. The image transformation neural network modifies a corresponding source image in the source image sequence by computing motion estimation differences between the identified driving image and the identified expression image.” – confirming that two source images (i.e., driving image and expression image) and an image of the source image sequence (reference frame of the reference video in claim 1) is used to generate modified source image, where features of pose and expression are combined in the individual generated image to imitate features of an individual reference frame of the source image sequence.
Applicant argues – the image transformation neural network modifies a corresponding source image in the source image sequence by computing motion estimation differences between the identified driving image and the identified expression image, and thereafter concludes that only one source image is used to modify each image in the source image sequence. The conclusion is misconstrued, and incorrect. According to the arguments provided above in points 1-3, it is clear that driving image and expression image are identified which are used to modify each image in the input image sequence such that features of pose and expression are combined in the modified image to imitate similar features in the reference frame of each input image of the input image sequence. Demyanov’s disclosure “The image transformation neural network modifies a corresponding source image in the source image sequence by computing motion estimation differences between the identified driving image and the identified expression image.” – should not be interpreted as a single image being used only in the modification, rather the understanding should be what the text means in plain reading. It is clear from the line that (a) corresponding source image in the source image sequence, (b) driving image and (c) expression image are used in the neural network to generate the modified source image. The estimation of motion difference is a means of achieving such modification. Examiner contends that instant application also does similar motion transfer to the modified image (see applicant’s specification ¶0060, 0064… etc. from PGPUB, US 20240095989 ).
For further details see the Office Action below.
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 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.
(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.
Claims 1-20 are rejected under 35 U.S.C. 102(a)(1) and/or 102(a)(2)as being anticipated by Demyanovet al. (US 11645798 B1, hereinafter Demyanov).
Regarding claim 1, Demyanov discloses a processor (1006, fig. 10), comprising:
one or more circuits (figs. 1-3, 7, 10) to use one or more neural networks (fig. 7) to generate one or more images (step 910, fig. 9) of one or more objects (head, face of a human, fig. 8) based, at least in part, on one or more reference frames of a reference video (corresponding source image in the source images, col. 21, line 57) and, at least two source images of the one or more objects (identified driving image and expression image of steps 904-906, fig. 9), wherein an individual image of the one or more generated images is generated (modifies a corresponding source image in the source image, col. 21, line 57) based, at least in part, on a modifying different features of the one or more objects from different images of the at least two source images to be combined in the individual generated image (Col. 21, lines 25-27, Demyanov discloses - “At operation 904, the facial animation system 700 identifies driving image sequence data to modify face image feature data in the source image sequence.” – confirming that driving image data indeed modifies face image features in the source image sequence.
In Col. 21, lines 25-27, Demyanov discloses – “At operation 906, the facial animation system 700 identifies an expression dataset to modify face image feature data in the source image sequence” – confirming that driving image data indeed modifies face image features in the source image sequence) to imitate features of an individual reference frame of the one or more reference frames of the reference video (…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
Modified source image sequence, is generated and stored in steps 908 & 910 using a source image sequence and a driving image and an expression image [understood as two source images in recited limitation] at steps 902 and 904, fig. 9. See fig. 7 and abstract also.
In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 2, Demyanov discloses the processor of claim 1, wherein the at least two images comprise a first image of an object and a second image of the object and wherein the one or more circuits are to generate a generated video using the at least two images of the object to be included in the generated video by selecting a first feature of the object from the first image and a second feature of the object from the second image to include in one of the one or more images generated by the one or more neural networks as a generated frame in the generated video (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 3, Demyanov discloses the processor of claim 1, wherein the one or more neural networks are to generate a generated video by at least selecting a portion from a first image of the at least two source images and a portion from a second image of the al least two source images to generate one of the one or more images to include in the generated video, wherein the first and second images are different (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 4, Demyanov discloses the processor of claim 1, wherein the one or more circuits are to:
animate an object (Embodiments of the present disclosure generally relate to facial animation. More particularly, but not by way of limitation, the present disclosure addresses systems and methods for facial animation transfer under a wide range of head poses and facial expressions, Col. 1, lines 12-16) in the at least two source images with the highest resemblance to one or more objects in the reference video (Shapes can be represented as vectors using the coordinates of the points in the shape. One shape is aligned to another with a similarity transform (allowing translation, scaling, and rotation) that minimizes the average Euclidean distance between shape points. The mean shape is the mean of the aligned training shapes, Col. 9, lines 59-64); and
generate a video to comprise the animated object to mirror motion of the one or more objects in the reference video (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 5, Demyanov discloses the processor of claim 1, wherein the one or more circuits are to:
identify one or more features from the at least two source images (…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract.
For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54);
select a portion from the at least two source images comprising a feature that matches a feature identified in the corresponding individual reference frame of the same one or more reference frames of the reference video (…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence); and
generate a video to comprise an animated feature of the selected portion (The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33.
Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 6, Demyanov discloses the processor of claim 1, the one or more neural networks to generate a video comprising animated objects (The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33
Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 7, Demyanov discloses the processor of claim 1, wherein the one or more circuits are to use one or more neural networks to:
identify one or more features from each of the at least two source images; and
select a subset of the one or more features that match the individual reference frame of the one or more reference frames corresponding to the reference video to generate the generated video (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67, abstract.
Also see Col 2, lines 45-56. Col 4, lines 36-42. Col 8, lines 30-48. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 8, Demyanov discloses a system (abstract, figs. 2, 7, 10), comprising: one or more processors (1006, fig. 10) to use one or more neural networks (Abstract) to generate one or more images of one or more objects based, at least in part, on one or more reference frames of a reference video and at least two source images of objects, wherein an individual image of the one or more generated images is generated based, at least in part, on a corresponding individual reference frame of the one or more reference frames of the reference video and different features of the one or more objects from different images of the at least two source images (see substantively similar claim 1 rejection above).
Regarding claim 9, Demyanov discloses the system of claim 8, wherein the one or more processors are to cause one or more objects in the at least two source images to be animated based, at least in part, on matching features identified in the at least two images with features identified in the reference video (…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract. For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54.
The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33.
In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 10, Demyanov discloses the system of claim 8, wherein the one or more processors are to select different portions from the at least two source images to generate a frame of a video based, at least in part, on the selected different portions (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
(…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract.
For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54.
The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 11, Demyanov discloses the system of claim 8, wherein the one or more processors are to:
identify a feature in a first image and the feature in a second image (…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract.
For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54);
select the feature from the first or second images with the higher similarity score to a corresponding feature in an individual reference frame of the reference video (Shapes can be represented as vectors using the coordinates of the points in the shape. One shape is aligned to another with a similarity transform (allowing translation, scaling, and rotation) that minimizes the average Euclidean distance between shape points. The mean shape is the mean of the aligned training shapes, Col. 9, lines 59-64); and
generate a frame of video based, at least in part, on the selected feature (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 12, Demyanov discloses the system of claim 8, wherein the one or more processors are to generate a generated frame of generated video by using a portion of a first image and a portion of a second image, wherein the portion of the first image comprises a feature that is different from the feature in the portion of the second image (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 13, Demyanov discloses the system of claim 8, wherein the one or more processors are to animate an object identified in the at least two source images by selecting different portions from the at least two images to generate a frame of a generated video comprising the animated object (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 14, Demyanov discloses the system of claim 8, wherein the one or more processors are to use one or more neural networks to identify features and select different portions from the at least two images to generate a generated frame of a generated video (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
(…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract.
For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54.
The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33.
Also see Col 2, lines 45-56. Col 4, lines 36-42. Col 8, lines 30-48.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract)
Regarding claim 15, although wording is different, the material is substantively similar to the independent claim 1 above.
Regarding claim 16, Demyanov discloses the method of claim 15, wherein the at least two source images comprise a first image of an object and a second image of the object and wherein the generating comprises generating a frame of a generated video by selecting different features from the at least two source images based (In some examples, when a particular modification is selected along with content to be transformed, elements to be transformed are identified by the computing device, and then detected and tracked if they are present in the frames of the video. The elements of the object are modified according to the request for modification, thus transforming the frames of the video stream. Transformation of frames of a video stream can be performed by different methods for different kinds of transformation, Col. 8, lines 59-67. Also see fig. 7 and Abstract.
…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract.
For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54.
The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using motion estimation differences between the identified driving image and identified expression image, Col. 17, lines 52-67.
The image transformation neural network is trained to identify, for each image in the source image sequence, a driving image from the driving image sequence data and an expression image from the expression dataset. The driving image is identified based on the driving image having a similar head pose image to the image in the source image sequence. The expression image is identified based on the expression image having a similar head pose and a similar expression to the image in the source image sequence, Col. 21, lines 48-56. See figs. 6-7. Read Abstract).
Regarding claim 17, Demyanov discloses the method of claim 15, further comprising: identifying features from the at least two source images (…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence, Abstract.
For example, in some embodiments, features are located using a landmark, which represents a distinguishable point present in most of the images under consideration, Col. 9, lines 50-54);
comparing the features from the at least two source images with features in the corresponding individual reference frame of the one or more reference frames of the reference video; and
selecting portions from the at least two source images to generate a generated video based, at least in part, on the comparison (…the source image sequence comprising a plurality of source images depicting a head and face, identifying driving image sequence data to modify face image feature data in the source image sequence. The present disclosure addresses facial animation transfer under a wide range of head poses and facial expressions. An expression matcher system generates a source image sequence depicting a head and a face. The expression matcher system identifies driving image sequence data to modify face image feature data in the source image sequence, Col. 2, lines 27-33. Also see fig. 7 and Abstract.
…for each source image in the source image sequence provided by the source image data module 610, identifies a driving image from the animation data 602 based on the driving image having a similar head pose to the source image and identifies an expression image from the animation data 602 based on the expression image having a similar expression and head pose as depicted in the source image. The trained machine learning technique module 608 provides the identified driving image and identified expression image for each corresponding source image to the pose and expression match module 612. The pose and expression match module 612 modifies the corresponding source image in the source image sequence by using moti