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 .
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 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 of this title, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 2, 7, 9, 10, 15 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Agarwal et al. (US 20220051479 A1), referred herein as Agarwal in view of YUAN et al. US 20230162481 A1), referred herein as YUAN and Zeng et al. (US 20250166237 A1), referred herein as Zeng.
Regarding Claim 1, Agarwal in view of YUAN and Zeng teaches a method for rapidly digitally rendering apparel designs, comprising (Agarwal [0003] systems and methods for creating patterns on garments):
receiving a draft file of a first image comprising a sketch of one or more apparel items (Agarwal [0076] designer input 407 may be digital file including a wear pattern for a garment that was created by an artist, designer, or other individual… Designer input 407 may include line sketches, labelled sketches, text descriptions, reference garments, edited reference garments, such as masked reference garments or local patch edited garments; [0077] The designer input 407 can include a variety of sources from photographs to sketches to text description. The designer input 407 may include masked or patched images of reference garments),
receiving a text description comprising a description of the one or more apparel items (Agarwal [0077] The designer input 407 can include a variety of sources from photographs to sketches to text description. The designer input 407 may include masked or patched images of reference garments; [0006] the at least one ML model may include a first ML configured to indicate visual design elements based on the interpreted instructions, also referred to as text-to-image synthesis, and a second ML model configured to perform neural style transfer to combine the visual design elements and apparel content to generate the apparel design),
wherein the text description comprises a positive description of one or more attributes to be present in the one or more apparel items Agarwal [0024] the interpreted user instructions may include vectors based on the text data and/or recognized entities within the text data, and the at least one ML model may be configured to convert or map the vectors and/or the recognized entities to images of respective apparel design elements and to generate the apparel design that incorporates the apparel design elements);
Agarwal does not but YUAN teaches
and a negative description of one or more attributes to be excluded from the one or more apparel items (YUAN [0041] To mitigate the negative effect from augmented prompts, the training may separated into two stages. In a first stage, augmented texts are used for training. In a second stage, all augmented data is excluded for continuing training. The first stage may use all or some of the raw, non-augmented texts for training).
Agarwal in view of YUAN further teaches
selecting a text-image relation model configured to relate the text description to one or more latent representations of the text-image relation model (YUAN [0022] To learn from image-text pairs 217, such as the noisy web-scale data curated by data curation engine 205, pre-training model 212 may comprise a two-tower architecture including an image encoder 220 and a language encoder 222; [0079] finding a candidate word may include searching the shared latent space based on a vector encoded by the audio model for an audio input, in order to find a candidate word vector for decoding with the word model; Agarwal [0004] train a design model on a relationship between non-computer-generated garment design information and non-computer-generated garment wear patterns);
selecting a generative image model configured to generate an output image constrained by both the first image of the one or more apparel items and a first set of latent representations that includes at least one of the one or more latent representations of the text-image relation model (Agarwal [0077] The images may be synthesized using generative model 431... generative model 431 has been trained on a database of images of existing distressed garments; [0079] styleGAN can be made to receive conditional input by separately training encoders that map inputs (such as sketches or text) into the styleGAN latent space; [0075] wear pattern 401 may be a wear pattern corresponding to a garment that was generated by a computer. Wear pattern 401 may include wear pattern elements such as a location of the wear pattern on the garment, one or more shapes of the wear pattern, and a distress factor of the wear pattern; FIG. 4.401 Model Output);
parameterizing the generative image model with a set of model parameters (Agarwal [0090] the design input 307 can incorporate multiple forms of designer input 307 in order to integrate images, line drawings and text descriptions to create multi-parameter computer generated images 301).
Agarwal does not but Zeng teaches
comprising at least one of (i) a convolution step value, (ii) a constraint divergence value, (iii) a quality value of the output image; (iv) a size value of the output image; (v) a generative seed, (vi) a sampler model, and (vii) a decoding algorithm, and (viii) a denoise algorithm (Zeng [0064] causes a decoding portion trained to associate features with corresponding keywords from text prompts (e.g., by associating features of the dog with the word “dog,” and associating features of snow with “snow”) by causing the decoding portion to decode the features identified by the decoder portion; [0066] a data structure including image features identified in a convolution of an input image, and generates a new feature map that includes assigned weights to features that are common to a (same) subject in the images; [0082] a joint distribution can include a combined distribution of text features and image features such that said distribution can be sampled and said samples can be used to denoise and generate same objects (e.g., same dog) in different backgrounds according to text prompt inputs);
Agarwal in view of YUAN and further teaches
outputting the first set of latent representations from the text-image relation model (Agarwal [0084] Generative model 331 can generate multiple computer-generated wear patterns 301 such that a user may wish to compile 312 computer generated wear patterns 301 into a design collection 305. Design collections enable classifications of computer-generated wear patterns 301 into curated collections, such as design connection 405, where the wear patterns 301 are collected so as to facilitate the production of distressed garments);
inputting into the generative image model a set of inputs of the generative image model comprising (i) the draft file comprising the first image of the one or more apparel items (Agarwal [0076] designer input 407 may be digital file including a wear pattern for a garment that was created by an artist, designer, or other individual), (ii) the first set of latent representations of the text-image relation model (YUAN [0022] To learn from image-text pairs 217, such as the noisy web-scale data curated by data curation engine 205, pre-training model 212 may comprise a two-tower architecture including an image encoder 220 and a language encoder 222), and (iii) the set of model parameters (Agarwal [0090] the design input 307 can incorporate multiple forms of designer input 307 in order to integrate images, line drawings and text descriptions to create multi-parameter computer generated images 301); and
generating a first rendering file comprising a second image of the one or more apparel items modified by the text description Agarwal [0036] the ML models 138 include at least a first ML model and a second ML model. The first ML model may be configured to generate outputs that indicate one or more visual design elements (and/or one or more apparel contents) based on the output of the natural language processor 134 (e.g., one or more vectors, indication of recognized named entities, or a combination thereof), and the second ML model may be configured to generate the apparel design based on the one or more visual design elements (and/or the one or more apparel contents); Zeng [0130] In at least one embodiment, neural network training includes generating 808 one or more images (e.g., images 305a-305n) of the subjects of subject set 203 over one or more backgrounds described by the one or more background prompts 303a-303n. In at least one embodiment, neural network 304 is one or more of a text to image diffusion model, a latent text to image diffusion model, a stable diffusion inpainting model, or other neural networking model trained to connect text with images described by the text; [0511] In at least one embodiment, GPE 3310 includes a 3D pipeline 3312 for performing 3D operations, such as rendering three-dimensional images and scenes using processing functions that act upon 3D primitive shapes (e.g., rectangle, triangle, etc.)).
“to allow for at least one of rapid visualization, prototyping, and construction of the one or more apparel items” isn’t given any patentability weight due to the language of intended use.
YUAN discloses pre-training of computer vision foundational models. YUAN is analogous to the present patent application.
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Agarwal to incorporate the teachings of YUAN, and applying the negative description and pairing models into the method for facilitating automated apparel design using deep learning techniques.
Doing so would provide a computer vision foundation model that is pre-trained via unified image-text contrastive learning.
Zeng discloses methods for using neural networks for generating multiple related images. Zeng is analogous to the present patent application.
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Agarwal to incorporate the teachings of Zeng, and applying the output weight and/or input/output data, and/or other parameters to configure neurons or layers of a neural network into the method for facilitating automated apparel design using deep learning techniques.
Doing so would improve neural networks that generate images as well as ways to improve training of these neural networks.
Regarding Claim 2, Agarwal in view of YUAN and Zeng teaches the method of claim 1, and further teaches further comprising:
generating a second set of model parameters, wherein the second set of model parameters differs from the set of model parameters in at least one of: the constraint divergence value, the generative seed, the text description, the positive description, the negative description, and the sampler model (Agarwal [0064] The design improver 208 may be configured to perform operations to refine the apparel design, such as providing information to the design generator 206 to be used as training data to train the ML models to generate additional apparel designs with improved likelihood of matching user expectations);
generating a second set of latent representations of the one or more latent representations of the text-image relation model based on reduced constraint in the of the text-image relation model (Agarwal [0078] With latent space additions, a latent vector z may be used to interpolate new instances of image representations. The discriminator 312 may be configured to predict whether image and text pairs match or not. Additionally, the images from interpolated text embeddings can fill in the gaps in the data manifold that are present during training);
generating an alternative rendering file comprising a third image of the one or more apparel items (Agarwal [0033] Generating the context-independent signatures may include iterating through all words or phrases of the embeddings vocabulary, selecting the context-independent signature for each term that satisfies a threshold and treating the selected term; FIG. 6.608-610);
transmitting the rendering file and the alternative rendering file to a user (Agarwal [0039] the virtual changing room manager 140 may be configured to provide the image data from the user device 102 and data indicating the apparel design generated by the design generation engine 136 as inputs to one or more ML models);
receiving a preference selection from the user for the alternative rendering file (Agarwal [0042] [0042] Additionally or alternatively, the user may perform one or more gestures to indicate selected controls, and the user device 102 transmits image data captured by the camera 114 to the server 120. For example, the gestures may include an “okay” gesture or a “thumbs up” gesture indicating approval of the apparel design); and
storing the second set of model parameters and optionally locking the generative seed (Agarwal [0086] the server 120 may transmit data indicating the apparel design to the user device 102, store the data indicating the apparel design at the memory 124, or initiate a transaction based on the apparel design).
Regarding Claim 7, Agarwal in view of YUAN and Zeng teaches the method of claim 1, and further teaches further comprising:
authenticating at least one of a generative user and a device of the generative user (Zeng [0372] upon receiving a hypervisor call, hypervisor 2096 verifies that operating system 2095 has registered and been given authority to use graphics acceleration module 2046); and
determining the generative user is authorized to access at least one of (i) two or more image files each comprising one or more apparel elements; (ii) the text description of the one or more apparel items; and (iii) an adaptation tuning model comprising a model trained on the two or more image files each comprising the one or more apparel elements (Zeng [0617] access to APIs in cloud 4226 may be restricted to authorized users through enacted security measures or protocols. In at least one embodiment, a security protocol may include web tokens that may be signed by an authentication (e.g., AuthN, AuthZ, Gluecon, etc.) service and may carry appropriate authorization),
wherein inputs to the generative image model further comprising the adaptation tuning model (YUAN [0021] Curated data may be used to develop computer vision foundation model 210. Computer vision foundation model 210 may be divided into two parts. One part is a pre-training model 212 (shown on the left side). The other part is a set of adaptation models 215 (shown on the right side). The adaptation models extend the capabilities of the pre-trained model so the foundation model can support two or more different tasks).
Regarding Claims 9, 10 and 15, Agarwal in view of YUAN and Zeng teaches a device for rapid apparel prototyping and design visualization, the device comprising: a computer comprising: one or more processors; a memory (Agarwal [0003] systems and methods for creating patterns on garments; FIG. 1). The metes and bounds of the claims substantially correspond to the limitations set forth in claims 1, 2 and 7; thus they are rejected on similar grounds and rationale as their corresponding limitations.
Regarding Claim 17, Agarwal in view of YUAN and Zeng teaches a system for rapid apparel prototyping and design visualization, the system comprising: a coordination server comprising: a processor of the coordination server; a memory of the coordination server, and a network communicatively coupling the coordination server and the generative server (Agarwal [0003] systems and methods for creating patterns on garments; FIG. 1). The metes and bounds of the claims substantially correspond to the limitations set forth in claim 1; thus they are rejected on similar grounds and rationale as their corresponding limitations.
Claim(s) 6 and 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Agarwal et al. (US 20220051479 A1), referred herein as Agarwal in view of YUAN et al. US 20230162481 A1), referred herein as YUAN, Zeng et al. (US 20250166237 A1), referred herein as Zeng and Pakhomov et al. (US 20240135514 A1), referred herein as Pakhomov.
Regarding Claim 6, Agarwal in view of YUAN and Zeng teaches the method of claim 1, but does not teach all the claimed limitations herein.
In viewing of Pakhomov, the prior art teaches further comprising: inputting at least one of the draft file and the first rendering file into a computer vision model trained with a set of training images comprising data designating a material value associated with at least one of an apparel item and an apparel element (Pakhomov [0315] the scene-based image editing system 106 utilizes low-level feature maps to accurately predict attributes such as, but not limited to, colors (e.g., red, blue, multicolored), patterns (e.g., striped, dotted, striped), geometry (e.g., shape, size, posture), texture (e.g., rough, smooth, jagged), or material (e.g., wooden, metallic, glossy, matte) of a portrayed object; [0357] FIG. 19A shows that the attribute menu 1910 provides object attributes indicators 1912a-1912c, indicating the shape, color, and material of the object 1908, respectively).
Pakhomov discloses method that modify digital images via multi-layered scene completion techniques facilitated by artificial intelligence. Pakhomov is analogous to the present patent application.
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Agarwal to incorporate the teachings of Pakhomov, and material values input from a user into the method for facilitating automated apparel design using deep learning techniques.
Doing so would provide a method with flexible and intuitive editing of digital images while efficiently reducing the user interactions typically required to make such edits.
Regarding Claim 14, Agarwal in view of YUAN and Zeng teaches the device of claim 9. The metes and bounds of the claims substantially correspond to the limitations set forth in claim 6; thus they are rejected on similar grounds and rationale as their corresponding limitations.
Allowable Subject Matter
Claim(s) 3-5, 8, 11-13, 16 and 18-20 is/are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter:
Regarding Claim 3, Agarwal in view of YUAN and Zeng teaches the method of claim 1, but does not teach all the claimed limitations herein.
In viewing of Pakhomov, the prior art teaches further comprising:
generating a graph data structure comprising (Agarwal [0033] treating the selected term in the trail of a word as a complete graph where edge strengths correspond to distance values, and selecting a node having the maximum strength of all the nodes):
a model selection node comprising attributes and associated values specifying a directed edge from the model selection node to an image generation model node , a directed edge from the model selection node to a text-image relation node, and a directed edge from the model selection node to a decoding algorithm node (Pakhomov [0137] the scene-based image editing system 106 selects the object segmentation machine learning model 310 based on the object labels of the object identified by the object detection machine learning model 308. Generally, based on identifying one or more classes of objects associated with the input bounding boxes, the scene-based image editing system 106 selects an object segmentation machine learning model tuned to generate object masks for objects of the identified one or more classes; [0138] an object mask includes a segmentation boundary indicating a predicted edge of one or more objects as well as pixels contained within the predicted edge; [0239] [0239] Additionally, as shown in FIG. 10, the image analysis graph 1000 associates characteristic attributes with one or more of the nodes 1004a-1004g to represent characteristic attributes of the corresponding characteristic categories; [0240] As further shown in FIG. 10, the image analysis graph 1000 associates characteristic attributes with one or more of the edges 1006a-1006h to represent characteristic attributes of the corresponding characteristic relationships represented by these edges 1006a-1006h);
the text-image relation node comprising attributes and associated values specifying a directed edge from the model selection node to the image generation model node and storing at least one of (i) the positive description of the one or more apparel items and a directed edge to the generative image model, and (ii) the negative description of the one or more apparel items (Pakhomov [0240] the image analysis graph 1000 associates a characteristic attribute 1012b with the edge 1006h indicating that at least some objects portrayed in a digital image have relationships with one another);
However, the prior art does not teach
a base parameterization node comprising attributes and associated values storing the size value of the output image and at least one of a directed edge from the base parameterization node to the image generation model node and from the image generation model node to the base parameterization node; and
the image generation model node, comprising attributes and values storing the set of model parameters and a directed edge from the image generation model node to the decoding algorithm node.
Therefore, the claim 3 in context of the claim 1 as a whole would be allowable if rewritten in independent form.
Regarding Claim 4, Agarwal in view of YUAN and Zeng teaches the method of claim 1, but does not teach all the claimed limitations herein.
In viewing of Pakhomov, the prior art teaches further comprising:
inputting at least one of the draft file and the first rendering file into a computer vision model trained with a first set of training images comprising data distinguishing apparel items (Agarwal [0006] the at least one ML model may include a first ML configured to indicate visual design elements based on the interpreted instructions, also referred to as text-to-image synthesis, and a second ML model configured to perform neural style transfer to combine the visual design elements and apparel content to generate the apparel design);
generating a segmentation file comprising one or more boundary designations of the one or more apparel items (Pakhomov [0131] an approximate boundary includes at least a portion of a detected object and portions of the digital image 316 not comprising the detected object. An approximate boundary includes various shape, such as a square, rectangle, circle, oval, or other outline surrounding an object. In one or more embodiments, an approximate boundary comprises a bounding box; [0132] the detection-masking neural network 300 generates segmentations masks that better define the boundaries of the object);
receiving a selection of a boundary designation of the one or more boundary designations of the one or more apparel items to select an apparel item bounded by the boundary designation (Pakhomov [0136] the object segmentation machine learning model 310 corresponds to one or more deep neural networks or models that select an object based on bounding box parameters corresponding to the object within the digital image 316).
However, the prior art does not teach
receiving a new text description comprising a description of the apparel item bounded by the boundary designation;
outputting a second set of latent representations from at least one of the text-image relation model and a different text-image relation model based on the new text description;
inputting into the generative image model at least a portion of the first rendering file within the boundary designation, the second set of latent representations of the text-image relation model, and a different set of model parameters; and
generating a second rendering file re-rendering the apparel item bounded by the boundary designation.
Therefore, the claim 4 in context of the claim 1 as a whole would be allowable if rewritten in independent form.
Regarding Claim 5, Agarwal in view of YUAN and Zeng teaches the method of claim 4. Claim 5 would be allowable by virtues of its dependency.
Regarding Claim 8, Agarwal in view of YUAN, Zeng and Pakhomov teaches the method of claim 6. REPHAELI further teaches further comprising:
wherein the set of inputs of the generative image model further comprising at least one of the line map file, the depth map file, and the color map file (REPHAELI [0041] User interface 130 may include a display 132 and controls 134. In an example embodiment, the user interface 130 may provide a way for a user to interact with the system 100. Display 132 may include an LED display configured to provide information indicative of distances to certain objects within the field of view. For example, the display 132 may provide a live view representation of the field of view of the image sensor 120. Furthermore, a depth map may be overlaid on the live view representation or displayed in a side-by-side view. The depth map may include one or more of: numerals or text indicating absolute distance units (e.g. “2.53 mm”), a color-mapped representation of the distance to various regions or objects within the field of view (e.g., a “heat” or rainbow-color map), or a topographical-style line map indicating iso-distance or pseudo-iso-distance features. Other ways to provide distance information via display 132 are contemplated. Alternatively or additionally, display 132 may be operable to provide notifications, information, and/or options to a user of system 100).
However, the prior art does not teach
inputting the draft file into a linage mapping model outputting a line map file of at least one of material-material boundaries of the one or more apparel items, material-skin transition boundaries, and material-background boundaries of the one or more apparel items;
inputting the draft file into a depth mapping model outputting a depth map file designating a perceived depth of at least one of (i) a first apparel item of the one or more apparel items relative to a second apparel item of the one or more apparel items; and (ii) a first material of the one or more apparel items relative to a second material of the one or more apparel items; and
inputting the draft file into a color mapping model outputting a color map file designating a color of at least one of (i) the first apparel item relative to the second apparel item; and (ii) the first material relative to the second material.
Therefore, the claim 8 in context of the claims 1 and 6 as a whole would be allowable if rewritten in independent form.
Regarding Claims 11-13 and 16, Agarwal in view of YUAN and Zeng teaches the device of claim 9. The metes and bounds of the claims substantially correspond to the limitations set forth in claims 3-5 and 8; thus they are objected for the same reason as above.
Regarding Claims 18-20, Agarwal in view of YUAN and Zeng teaches the system of claim 17. The metes and bounds of the claims substantially correspond to the limitations set forth in claims 3-5 and 8; thus they are objected for the same reason as above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
REPHAELI et al. (US 20170347043 A1), referred herein as REPHAELI.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Samantha (Yuehan) Wang whose telephone number is (571)270-5011. The examiner can normally be reached Monday-Friday, 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, King Poon can be reached on (571)272-7440. 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.
/Samantha (YUEHAN) WANG/
Primary Examiner
Art Unit 2617