DETAILED ACTION
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Gebre (US 12,033,254) in view of Kim (US 12,354,195).
Referring to claims 1, 9 and 17, Gebre discloses a system comprising: a processor (fig. 8, processor 802); a memory component (fig. 8, memory 806) storing instructions cause the processor to:
accessing an image (fig. 3, original face image 302) representing a source image (fig. 3, camera captured face images; 5:44-46);
generating a training dataset (fig. 3, computer generated image 332) representing a target image (fig. 2, image with various hair styles);
training a base generative neural network (fig. 3, first trained neural network 306; fig. 1, input to neural network model 102) to generate images representing the source image domain (fig. 1, generate face layer 104 images) and images representing adjacent source image domains (fig. 1, generate hair layer 104 images);
training a final generative neural network (fig. 3, discriminative networks 330/336; 9:4-7, neural network), using the base generative neural network (fig. 3, first trained neural network 306) and the training dataset (fig. 3, computer generated image 332);
generating a image dataset (9:7-24, acceptable generated face layer and hair layer images with corresponding actual image) using the final generative neural network;
training an image generation neural network (fig. 4, second trained neural network 402; fig. 1, neural network model 106), using the paired image dataset (fig. 1, such as hair layer; 6:40-59), to generate a modified image (fig. 1, generate modified hair layer 108) for an input image (fig. 1, input image 102; fig. 3, input image 302); and
generating a modified image by applying the image generation neural network to the accessed image (fig. 1, generate modified image 110; 6:60-7:29, combine modified hair layer with face layer), the modified image representing the target image domain (fig. 2, one of the faces with hair).
Kim discloses
a source image domain (fig. 1, source image),
a target image domain (fig. 1, target image),
a paired image dataset (8:45-52, pairs between sources and target), and
wherein applying the image generation neural network (fig. 4, StyleGAN2) transforms the accessed image (fig. 1, result image) from the source image domain to the target image domain (fig. 2, transform from source image and target image into result image similar to the target image),
the source image domain being distinct from the target image domain (fig. 2, source image different from target image).
Gebre and Kim are analogous art because they are from the same field of endeavor in using neural network modifying image. Before the time of the filing, it would have been obvious to a person of ordinary skill in the art, having the teaching of Gebre and Kim before him or her to modify the source/target images in the neural networks of Gebre to include the source/target/result images in StyleGAN of Kim, thereafter a target-image alike result-image is generated from a source image with features from target image using machine learning networks. The suggestion and/or motivation for doing so would be obtaining the advantage of high-quality images (4:44-48) as suggested by Lee. Therefore, it would have been obvious to combine Gebre with Lee to obtain the invention as specified in the instant application claims.
As to claims 2 and 10, Gebre discloses the system of claim 1,
wherein the training dataset comprises textual data and image data describing the target domain (11:49-53, training dataset images; fig. 3, annotation 320).
As to claims 3 and 11, Lee discloses the system of claim 1, wherein
the paired image dataset comprises a plurality of image pairs (8:45-52, possible pairs),
each image pair in the plurality of image pair comprising an original image (8:45-52, sources) corresponding to the source image domain and a stylized image (8:45-52, targets) corresponding to the target image domain.
As to claims 4 and 12, Gebre discloses the system of claim 1, wherein
the base generative neural network is used as initialization for training of the final generative neural network (fig. 1, first trained neural network 102 initialize training of the second trained neural network 106 etc.).
As to claims 5, 13 and 18, Gebre discloses the system of claim 1, wherein
training the base generative neural network comprises training the base generative neural network on an image dataset (fig. 2, face images with different hair layers), wherein
each image in the image dataset has a condition (6:48-59, hairstyles) representing an adjacent source domain of the image.
As to claims 6, 14 and 19, Gebre discloses the system of claim 1, wherein
neural network layers of the base generative neural network can accept a set of conditions (fig. 3, expressions 328/334) associated with the adjacent source domains.
As to claims 7, 15 and 20, Gebre discloses the system of claim 6, comprising:
applying a one-hot conditioning (fig. 3, expression 328/334 each represents a one-hot condition) to the base generative neural network; wherein each condition in the set of condition is represented as a vector (fig. 3, expressions 328/334); and
supplementing random gaussian noise (10:39-40, random noise) associated with the neural network layers of the base generative neural network with the set of conditions.
As to claims 8 and 16, Gebre discloses the system of claim 6, comprising:
modifying a vector (10:39-47, noise vector) representation of each neural network layer of the base generative neural network to incorporate data representing a respective condition in the set of conditions (10:39-47, modifying noise vector introduces variety into the modified hair layer).
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to examiner Cheng-Yuan Tseng whose telephone number is (571)272-9772, and fax number is (571)273-9772. The examiner can normally be reached on Monday through Friday from 09:00 to 17:30 Eastern Time. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alicia Harrington can be reached on (571)272-2330. The fax phone number for the organization where this application or proceeding is assigned is (571)273-8300. Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at (866)217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call (800)786-9199 (IN USA OR CANADA) or (571)272-1000.
/CHENG YUAN TSENG/Primary Examiner, Art Unit 2615
Prosecution Insights