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 .
Priority
The present application claims benefit of provisional Application No. 63/602,509 filed on 11/24/2023. Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Information Disclosure Statement
The information disclosure statement(s) (IDS) submitted on 11/06/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claims 9 and 19 are objected to because of the following informalities:
Regarding claims 9 and 19, lines 2-3 of both claims read “temporally omitting all of attention layers.” This phrase lacks proper grammatical form (i.e. “of” is grammatically improper) and appears inconsistent with the specification. The specification states both “the temporal attention layers are temporarily omitted” and “temporally omitting all of attention layers” (see ¶ [0046]).
Applicant may wish to amend the phrase to “temporarily omitting all temporal attention layers,” “temporarily omitting all attention layers,” “temporally omitting all
Claim Rejections - 35 USC § 112
Claims 15 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second
paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject
matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C.
112, the applicant), regards as the invention.
Claim 15 recites “the video animation,” see line 2. There is insufficient antecedent basis for “the video animation” because a video animation is not previously established.
Rejection is expected to be overcome by amending claim to be dependent of claim 14 (similar to mirrored computer system claim 5 being dependent of claim 4) because claim 14 establishes a video animation, or amending the claim to establish a video animation and its relationship to elements found in the independent claim 11 (similar to how “a video animation” is established in claim 14 with respect to independent claim 11).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness
rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or
nonobviousness.
Claims 1-3, 6, 8, 10-13, 16, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (“DisCo: Disentangled Control for Referring Human Dance Generation in Real World”; copy provided by Examiner; hereafter “Wang”) in view of Guo, Y. et al. (“AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning”; later version of reference provided by Applicant in IDS; copy of earlier version (prior to effective filing date of claimed invention) that’s relied upon for cited references in section below is provided by Examiner; hereafter “Guo”).
Regarding computing system claim 1 and computerized method claim 11,
Wang teaches: A computing system,
[claim 11: A computerized method], comprising: processing circuitry (Wang teaches a diffusion model computer system and computerized method, referred to as “DISCO,” for human image/video dance synthesis (see Abstract and last paragraph of p.2 titled “Model Architecture for Disentangled Control”), stating “latent diffusion Models and ControlNet…are the basis of DISCO” (p. 4, first paragraph) and “Latent Diffusion Models (LDM) is a type of diffusion model that operates in the encoded latent space of an autoencoder…An exemplary LDM is the popular Stable Diffusion (SD) which consists an autoencoder… and a time-conditioned U-Net…” (p. 4, section 3.1). Accordingly, Wang teaches a computer system and computerized method utilizing processing circuitry to execute a diffusion-based image generation framework.) configured to implement:
an appearance encoder configured to encode a reference image into an appearance embedding (Wang teaches receiving a reference human image and encoding the image into image embedding for conditioning the diffusion model, stating “Specifically, given a reference human image, we can first utilize an existing human matting method…” (p. 4, second paragraph of 3.2) and further teaches that “we adopt the local CLIP image embeddings right before the global pooling layer, for more fine-grained human semantics encoding. Consequently, the original text embedding ctext ∈ Rl× d is superseded by the local CLIP image embeddings of the human foreground” (p. 5 first paragraph; see FIG 2). Accordingly, under the broadest reasonable interpretation, the CLIP image encoder shown in Figure 2 reasonably teaches the claimed appearance encoder, as it encodes the reference human image into local CLIP image embeddings, which reasonably corresponds to the claimed appearance embedding.);
a pose control network configured to receive as input a target pose sequence and in response extract a motion condition from the target pose sequence (Wang teaches a pose ControlNet that receives a target pose as conditioning input and generates pose conditioning for the diffusion model (p. 4 second paragraph of 3.2 “Specifically, given a reference human image…Next, we explain how all three conditions, the human foreground f, the background g and the desired pose p, are incorporated into DISCO.”; p. 5 second paragraph “For pose p, we adopt the vanilla design of ControlNet. Specifically, we embed the pose image into the same latent space as the Unet input via with four convolution layers, and dedicate a ControlNet branch τθ to learn the pose control…. τθ and µθ represent the two ControlNet branches for pose condition p and background condition g, respectively.”). Accordingly, under the broadest reasonable interpretation, Wang’s ControlNet branch reasonably corresponds to the claimed pose control network, and the pose condition generated by the ControlNet branch reasonably corresponds to the claimed motion condition, as both are conditioning information derived from the input target pose supplied to the diffusion model to control motion generated.); and
a trained (Wang teaches a trained latent diffusion model by teaching the DISCO model is fine-tuned using the latent diffusion modeling objective and optimizes the denoising network during training (see p. 5, lines 1-2 of last paragraph of 3.2; p. 4 second paragraph of 3.1). Wang further teaches the trained diffusion model receives both the appearance embedding and pose condition for image generation, as described with respect to the previous limitation found above. Specifically, Wang teaches the diffusion model receives the appearance embedding generated from the reference human image together with the pose condition τθ(p) generated by the ControlNet branch, stating “the original text embedding ctext ∈ Rl× d is superseded by the local CLIP image embeddings of the human foreground” (p. 5 first paragraph), and that “τθ and µθ represent the two ControlNet branches for pose condition p and background condition g, respectively” (p. 5, last two lines of 3.2), thereby teaching the trained diffusion model receives both the appearance embedding and the conditioning information generated from the target pose. Under the broadest reasonable interpretation, the pose condition τθ(p) reasonably teaches the claimed motion condition.)
and generate a denoised animation sequence (Wang teaches generating an animation/video sequence from the reference image and pose sequences (p. 7, second paragraph, “Given a reference image and a target pose sequence…we generate the video frame-by-frame, with the reference image and a single pose as the inputs to DISCO.”). Wang further teaches the DISCO model uses a latent diffusion model trained to predict noise via the objective and explains the SD is trained to “learn the reverse denoising process” (p. 4, section 3.1, subsection “Latent Diffusion Models”).)
Wang fails to explicitly disclose: a trained video diffusion model including a temporal attention mechanism and generating the denoised animation sequence using a trained video diffusion model including a temporal attention mechanism, rather DISCO teaches its video generation is frame-by-frame. Wang expressly states the framework “can generate videos with surprisingly good temporal consistency, even without explicit temporal modeling” and identifies “explicit temporal modeling to improve temporal consistency” as future work (see p. 8 last two lines of 4.3; p. 9, second paragraph of Conclusion).
In a related art, Guo teaches: a trained video(Abstract “insert a newly initialized motion modeling module into the frozen text-to-image model and train it on video clips to distill reasonable motion priors.”; see Figure 2 and its accompanying description “our method first trains a motion modeling module on video datasets”; Guo teaches the motion modeling module utilizes “vanilla attention along the temporal dimension… to learn the proper motion priors” (see p. 2 right column, second paragraph).).
Guo further teaches: generating the denoised animation sequence [using a trained video diffusion model including a temporal attention mechanism] (Guo teaches the inserted motion modeling module is incorporated into the diffusion denoising process, wherein the base T2I model (e.g. Stable Diffusion) can produce “diverse and personalized animated images via iteratively denoise process” (see p. 4, Figure 2 and accompanying description). Guo further teaches that the motion modeling module is trained on video datasets and inserted into the diffusion model to “generate temporally smooth animation clips” (Abstract), thereby teaching generation of a denoised animation sequence using the trained video diffusion model including the temporal attention mechanism mentioned in Guo’s teachings corresponding to the previous limitation addressed above).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify the DISCO framework taught by Wang by incorporating the trained motion modeling module of Guo because Wang already teaches generating human dance videos from a reference image and pose sequence using a Stable Diffusion-based diffusion framework, but acknowledges (1) the lack of explicit temporal modeling and (2) the potential to improve temporal consistency through explicit temporal modeling (see Wang p. 8, last two lines of 4.3; p. 9, second paragraph of Conclusion), whereas Guo teaches incorporating a trained temporal motion modeling module into a Stable Diffusion-based diffusion model to learn motion priors from video datasets and improve temporal consistency. The modification would have predictable improved temporal consistency across successive generated video frames while preserving Wang’s existing diffusion framework.
Regarding claims 2 and 12,
Wang and Guo teach the computing system of claim 1 and the computerized method of claim 11.
Guo further teaches: wherein the video diffusion model includes a pretrained motion module including a transformer that has been trained on successive video frames to predict motion of image features in successive frames (Guo teaches a motion modeling module that is inserted into a diffusion model and separately trained on video data (Abstract “insert a newly initialized motion modeling module into the frozen text-to-image model and train it on video clips to distill reasonable motion priors”; p. 3 left column lines 2-3 “Our method is built upon a pre-trained text-to-image model”; p. 4 Figure 2 and accompanying description “our method first trains a motion modeling module on video datasets”; “Details of Motion Module…Our module is a vanilla temporal transformer” utilizing “vanilla attention along the temporal dimension… to learn the proper motion priors” (see p. 5 Figure 3 and its description, and p. 2 right column, second paragraph). Under the broadest reasonable interpretation, Guo’s motion modeling model reasonably teaches the claimed pretrained motion module, the vanilla temporal transformer reasonably teaches the claimed transformer, and the training on video datasets reasonably teaches training on successive video frames. Further, because the temporal transformer taught by Guo learns motion prior across successive frames, Guo reasonably teaches predicting motion of image features in successive frames.), and
the processing circuitry is further configured to implement a temporal video fusion operation using the pretrained motion module to generate the denoised animation sequence (Guo teaches the trained motion modeling module is incorporated into the diffusion model to generate animation through the diffusing process (Figure 2 description “the once-trained motion module can turn any personalized model tuned upon the base T2I model into an animation generator, then produce diverse and personalized animated images via iteratively denoise process”; Abstract “models generate temporally smooth animation clips”; p. 2 right column, second paragraph “vanilla attention along the temporal dimension”). Under the broadest reasonable interpretation, utilizing temporal attention to aggregate temporal information across successive video frames during the iterative denoising process reasonably teaches the claimed temporal video fusion operation using the pretrained motion module to generate the denoised animation sequence.).
Regarding claims 3 and 13,
Wang and Guo teach the computing system of claim 2 and the computerized method of claim 12.
Guo further teaches: wherein the pretrained motion module is trained by:
adding sinusoidal framewise positional encoding to successive frames of a training video to encode a position of each frame within the training video (p. 5 right column, lines 7-10, “we add sinusoidal position encoding to the self-attention blocks to let the network be aware of the temporal location of the current frame in the animation clip”); and
training the motion module to predict motion of visual features within the images of successive video frames by using a temporal attention mechanism (Guo teaches training the motion modeling module on video datasets (see Figure 2 and its accompanying description) and implementing the module using temporal transformers that utilize “vanilla attention along the temporal dimension” and “self-attention blocks operating along the temporal axis” to learn motion parts from successive video frames and “capture the temporal dependencies between features at the same location across the temporal axis” (p. 2, paragraph 2 of right column; p. 5 last paragraph of left column through end of paragraph found at top of right column). Under the broadest reasonable interpretation, learning motion priors across successive video frames using temporal self-attention reasonably teaches training the motion module to predict motion of visual features within successive video frames.)
that computes attention for each element of an image across elements in the successive frames of video using the framewise positional encoding (Guo teaches that temporal transformer utilizes “self-attention blocks operating along the temporal axis” to process the feature map and that the operation “captures temporal dependencies between features” across successive video frames whole utilizing the sinusoidal positional encoding discussed in the previous limitation (p. 5 last paragraph of left column through end of paragraph found at top of right column). Under the broadest reasonable interpretation, computing temporal self-attention across corresponding features of successive video frames using positional encoding reasonably teaches computing attention for each element of image across elements in successive frames of video.).
Regarding claims 6 and 16,
Wang and Guo teach the computing system of claim 1 and the computerized method of claim 11.
Wang further teaches: wherein the reference image includes an image of a human (Wang: see p. 2, second paragraph “Given a reference image designating a human subject…” and Figure 1).
Regarding claims 8 and 18,
Wang and Guo teach the computing system of claim 1 and the computerized method of claim 11.
Guo further teaches: the denoised animation sequence exhibits temporal consistency (Guo teaches its framework’s Stable Diffusion (SD) model executes the denoising process in the latent space of an autoencoder in order to preserve high visual quality images (see Guo p. 3, first paragraph of 3.1), the motion module is trained on video datasets and at inference the motion model produces animated images via an iteratively denoising process (see p. 4, Figure 2 and its accompanying description), and the method ultimately results in temporal consistency content (i.e. a denoised animation sequence) (see Figure 5 and p. 8 left column, third paragraph of 4.3 “our method generates temporally consistent content and maintains superior smoothness”).).
Regarding claim 10,
Wang and Guo teach the computing system of claim 1.
Wang further teaches: wherein a joint training using image datasets and video datasets is employed at a training time (Wang teaches training the DISCO framework (see Wang’s diffusion framework taught with respect to claim 1’s 103 rejection) using both image and video data. Specifically, Wang teaches that the Human Attribute Pre-training (HAP) stage involves “learning diverse human attributes from large-scale human image data,” and further teaches an “‘easy-to-hard’ training schema, with HAP focusing on reconstructing human images without pose editing, then learning pose control and implicit appearance distortions brough by motion in the fine-tuning stage” (see Wang p. 9, Pre-training Data Size paragraph). Wang further teaches video-based training, stating “During training, we sample the first frame of the video as the reference and all others at 30 fps as targets” (see p. 6, first paragraph of 4.1), thereby utilizing reference images and target video frames during diffusion model training. Under the broadest reasonable interpretation, training the same diffusion framework using image data during HAP and video data during the fine-tuning stage reasonably teaches joint training using image datasets and video datasets at training time.).
Claims 4 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (“DisCo: Disentangled Control for Referring Human Dance Generation in Real World”; copy provided by Examiner; hereafter “Wang”) in view of Guo, Y. et al. (“AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning”; copy provided by Examiner; hereafter “Guo”), and in further view of Schubert et al. (US 20140292642 A1; hereafter “Schubert”).
Regarding claims 4 and 14,
Wang and Guo teach the computing system of claim 1 and the computerized method of claim 11.
Wang further teaches: wherein denoised animation sequence is a video animation generated in multiple segments (Wang teaches generating a denoised animation sequence using a latent diffusion model trained according to a diffusion denoising objective (see Wang p. 4, section 3.1, subsection “Latent Diffusion Models) and generating a video from a reference image and target pose sequence, stating “Given a reference image and a target pose sequence… we generate the video frame-by-frame, with the reference image and a single pose as the inputs to DISCO” (p. 7, second paragraph). Under the broadest reasonable interpretation, generating the animation frame-by-frame reasonably teaches generating the video animation in multiple segments, each frame of the video constituting a segment of the generated animation.),
Wang fails to explicitly disclose: a sliding window technique has been applied to smooth transitions between segments during inference.
In a related art, Schubert teaches: a sliding window technique has been applied to smooth transitions between segments during inference (Schubert teaches “In a further step 114, a dynamic viewing area is defined. This is performed by a sliding-window-frame calculation” and further teaches “Defining the dynamic viewing area prepares for the virtual, location-based information to be displayed…as a sliding view” whereby “the virtual information follows the viewing direction seamlessly” and “smooth transitions are possible between the displayed virtual, location-based information” (see Schubert ¶ [0107]). Under the broadest reasonable interpretation, Schubert’s sliding-window-frame calculation is applied during display (i.e. inference/runtime operation rather than training) to provide seamless, smooth transitions between successive displayed portions of generated visual information, thereby reasonably teaching the claimed sliding window technique applied to smooth transitions between segments during inference.).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify the video generation framework of Wang, as further enhanced by Guo, to incorporate the sliding window technique taught by Schubert because Schubert teaches applying a sliding-window-frame calculation during runtime provides seamless and smooth transitions between successively displayed visual content. Doing so would have predictably improved transition smoothness between successive generated portions of the animation during inference while preserving Wang’s underlying diffusion-based generation framework. All three references are directed towards improving temporal continuity of sequentially presented visual information by utilizing information from successive frames or views to generate smoother visual output.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (“DisCo: Disentangled Control for Referring Human Dance Generation in Real World”; copy provided by Examiner; hereafter “Wang”) in view of Guo, Y. et al. (“AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning”; copy provided by Examiner; hereafter “Guo”), and in further view of Fu-Yun Wang et al. (“Gen-L-Video: Multi-Text to Long Video Generation via Temporal Co-Denoising”; copy provided by Examiner; hereafter “Fu-Yun”).
Regarding claim 15, as best understood based on the 35 U.S.C. 112(b) issue identified above,
Wang and Guo teach the computerized method of claim 11.
Wang in view of Guo further teaches: wherein the video animation is generated (Under the broadest reasonable interpretation of the claim, Wang and Guo’s teaching with respect to the generation of an animation sequence, found in 103 rejections for claims 1 and 11, reasonably teach generating a video animation. For further clarity, Wang teaches generating a video from a reference image and target pose sequence (see Wang p. 7, second paragraph, “Given a reference image and a target pose sequence… we generate the video frame-by-frame, with the reference image and a single pose as the inputs to DISCO”), and Guo further teaches generating temporally smooth animation using a trained video diffusion model including a temporal motion module (see Guo Figure 2 and corresponding description “produce diverse and personalized animated images via iteratively denoise process,” and Abstract “generate temporally smooth animation clips”).).
Wang and Guo fail to explicitly disclose: wherein the video animation is generated in multiple overlapping segments, and predictions for overlapping frames are averaged.
In a related art, Fu-Yun teaches: multiple overlapping segments (Fun-Yun teaches generating a video animation in multiple overlapping segments using a diffusion model, explaining that “Given a noisy video vt, they train a diffusion model pθ(vt−1|vt,c) with respect to noise prediction model ϵθ(vt,t,c),” and, “we consider the denoising process of the entire video as multiple short videos with temporal overlapping undergoing parallel denoising in the temporal domain” (see Fun-Yun p. 5, first two paragraphs of section 3.2). Fu-Yun further teaches projecting the original video into “short video segments,” where “S represents the stride among adjacent short video clips,” and teaches that “setting S to M//2 or M//4 yields excellent results and preserves efficiency. When setting S = M, our method degrades into isolated denoising” (p. 5, last paragraph). Under the broadest reasonable interpretation, the disclosed temporally overlapping short videos correspond to the claimed multiple overlapping segments, each performing diffusion-based predictions for the overlapping frames.).
and predictions for overlapping frames are averaged (Fu-Yun further teaches combining the predictions corresponding to the overlapping frames to determine the next denoised latent video, explaining, “the optimal
v
t
-
1
[i.e., the next denoised latent video] can be obtained by solving the following optimization problem,” and further teaches “for an arbitrary frame j in the video
v
t
-
1
… it should be equal to the weighted sum of all the corresponding frames in short videos that contain the j frame” (p. 6, third paragraph). Under the broadest reasonable interpretation, because each overlapping short video independently produces a diffusion-based prediction for the shared frame, the optimization determines the value of each overlapping frame from the collective contributions of the corresponding predictions generated by the overlapping short videos. A weighted summation (taught by Fu-Yun’s framework, as described above) is a form of averaging in which individual predictions contribute according to assigned weights rather than equally. Accordingly, Fu-Yun reasonably teaches averaging the predictions corresponding to overlapping frames.).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify the diffusion-based video generation framework of Wang, as modified by Guo, to utilize the temporally overlapping denoising technique of Fu-Yun because Fu-Yun recognizes that denoising an entire video as a whole “greatly limits the video length that they are able to generate” and that autoregression long video generation “suffers from severe content degradation” and “inference inefficiency” (see Fu-Yun p. 5, first paragraph of 3.2). Fu-Yun instead teaches denoising an entire video as “multiple short videos with temporal overlapping undergoing parallel denoising in the temporal domain” (see Fu-Yun p. 5, second paragraph of 3.2). Applying this known temporal denoising technique to the Wang and Guo framework would have predictably improved long video generation while reducing degradation and maintaining temporal continuity.
Claims 5 is rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (“DisCo: Disentangled Control for Referring Human Dance Generation in Real World”; copy provided by Examiner; hereafter “Wang”) in view of Guo, Y. et al. (“AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning”; copy provided by Examiner; hereafter “Guo”), and in further view of Schubert et al. (US 20140292642 A1; hereafter “Schubert”), and in further view of Fu-Yun Wang et al. (“Gen-L-Video: Multi-Text to Long Video Generation via Temporal Co-Denoising”; copy provided by Examiner; hereafter “Fu-Yun”).
Regarding Claim 5
Wang, Guo, and Schubert teach the computing system of claim 4. The remaining limitations in claim 5 mirror the limitations found throughout lines 1-3 of the computerized method of claim 15, shown to be taught by Wang, in view of Guo and Fu-Yun. For sake of brevity, refer to the teachings of Wang, Guo, and Fu-Yun and rationale to combine references detailed with respect to claim 15, as further applicable to the limitations of claim 5. Accordingly, claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Guo, Schubert, and Fu-Yun.
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (“DisCo: Disentangled Control for Referring Human Dance Generation in Real World”; copy provided by Examiner; hereafter “Wang”) in view of Guo, Y. et al. (“AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning”; copy provided by Examiner; hereafter “Guo”), and in further view of Wang, Yangtuanfeng et al. (US 20230123820 A1; hereafter “Yangtuanfeng”).
Regarding claim 20,
The limitations found on lines 5-10 of claim 20 mirror the scope of the limitations found in lines 3-10 of claim 1’s computer system and lines 2-7 of claim 11’s computerized method, previously shown to be taught by Wang and Guo. For sake of brevity, refer to the 103 rejections for claims 1 and 11 and Wang and Guos’ corresponding teachings with respect to the limitations found on lines 5-10 of claim 20.
Wang and Guo fail to explicitly teach: A non-transitory computer readable storage medium storing computer-executable instructions, wherein when executed by processing circuitry, the computer-executable instructions cause the processing circuitry configured to [perform the subsequent limitations found on lines 5-10 of claim 20].
In a related art, Yangtuanfeng teaches: A non-transitory computer readable storage medium storing computer-executable instructions, wherein when executed by processing circuitry, the computer-executable instructions cause the processing circuitry configured to [perform subsequent processing operations] (see Yangtuanfeng Abstract and ¶ [0158]-[0160]).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to store the video generation framework of Wang, as further enhanced by Guo, as executable instructions on a non-transitory computer readable storage medium, as taught by Yangtuanfeng, since such implementation merely involves storing known computerized methods/systems as executable instructions in a known medium, yielding predictable results, i.e. causing the processing circuity taught by Wang and Guo to perform the video generation techniques taught by Wang and Guo (refer to processing circuitry and subsequent configuration/implementation taught by Wang and Guo in the 103 rejections for mirrored claims 1 and 11). All references are directed to generating animated digital videos/clips informed by images and motion targeting with a goal of improving temporal coherency.
Allowable Subject Matter
Claims 7 and 17 are objected to as being dependent upon rejected base claims, but would be allowable if rewritten in independent form, including all of the limitations of the base claim and any intervening claims.
Claims 9 and 19 are objected to as being dependent upon rejected base claims, but would be allowable if the objections to the claims found in the “Claim Objections” section are resolved and the claims were rewritten in independent form, including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SAMUEL DAVID BAYNES whose telephone number is (571)272-0607. The examiner can normally be reached Monday - Friday 8:00 am - 5:00 pm.
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, Stephen R Koziol can be reached at (408)918-7630. 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.
/SDB/
Samuel D. Baynes
Examiner, Art Unit 2665
/Stephen R Koziol/Supervisory Patent Examiner, Art Unit 2665
Prosecution Insights