# US 12,182,911 B2 · Image generation method, image generation apparatus, and image generation system > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,182,911 B2 - Document ID: US12182911B2 - Status: Granted - Kind: B2 - Application number: 17/447,081 - Filed: 2021-09-08 - Published: 2021-12-23 - Granted: 2024-12-31 - Earliest claimed: 2019-03-11 - Assignees: Preferred Networks, Inc. - Applicants: Preferred Networks, Inc. - Inventors: Yanghua Jin; Huachun Zhu; Yingtao Tian - Web page: https://patentgrep.com/patent/US12182911B2 ## CPC - G06N 3/126: Evolutionary algorithms, e.g. genetic algorithms or genetic programming - G06F 18/25: Fusion techniques - G06N 3/0475: Generative networks - G06N 3/049: Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs - G06N 3/094: Adversarial learning - G06T 11/60: Creating or editing images; Combining images with text - G06T 13/40: of characters, e.g. humans, animals or virtual beings - G06N 3/045: Combinations of networks - G06N 3/047: Probabilistic or stochastic networks - G06N 3/065: Analogue means - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract An image generation apparatus includes at least one memory, and at least one processor configured to acquire first latent information of a first image and second latent information of a second image, generate fusion latent information by using the first latent information and the second latent information, and generate a fusion image by inputting the fusion latent information into a trained generative model. ## Claims ### Claim 1 An image generation apparatus comprising: at least one memory; and at least one processor configured to: acquire first latent information of a first image and second latent information of a second image, the first image including data related to a visual representation of a first object, and the second image including data related to a visual representation of a second object, generate fusion latent information by using the first latent information and the second latent information, and generate a fusion image by inputting the fusion latent information into a trained generative model, the fusion image including data related to a visual representation of a third object, wherein the at least one processor uses randomly determined information to generate the fusion latent information, wherein the first object, the second object, and the third object are respectively a first character, a second character, and a third character, the first character, the second character, and the third character being different from each other, wherein the visual representation of the first object indicates an appearance of the first character, the visual representation of the second object indicates an appearance of the second character, and the visual representation of the third object indicates an appearance of the third character, and wherein the first latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the first character, the second latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the second character, and the fusion latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the third character. ### Claim 2 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor is configured to perform a genetic operation on the first latent information and the second latent information to generate the fusion latent information. ### Claim 3 (depends on 2) The image generation apparatus as claimed in claim 2, wherein the genetic operation includes at least one of crossover, mutation, or selection. ### Claim 4 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor is configured to fuse elements of the first latent information and elements of the second latent information to generate the fusion latent information. ### Claim 5 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor is configured to select elements of the first latent information and elements of the second latent information to generate the fusion latent information. ### Claim 6 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor is configured to perform at least one of an arithmetic operation or a logical operation of the first latent information and the second latent information to generate the fusion latent information. ### Claim 7 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the first latent information includes main and sub information, and the second latent information includes main and sub information. ### Claim 8 (depends on 7) The image generation apparatus as claimed in claim 7, wherein the at least one processor is configured to: select one of the main and the sub information of the first latent information, select one of the main and the sub information of the second latent information, and generate the fusion latent information by using the selected information of the first latent information and the selected information of the second latent information. ### Claim 9 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor is configured to generate the fusion latent information that includes main and sub information. ### Claim 10 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the first latent information and the second latent information are represented in blockchain compliant codes. ### Claim 11 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the trained generative model is a generator trained in accordance with a generative adversarial network. ### Claim 12 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor displays, on a display device, the first image, the second image, the fusion image, and the information of the at least one of the hair style, the hair color, the eye color, the skin color, the expression, or the attachment of the third character. ### Claim 13 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the at least one processor displays, on a display device, the first image, the second image, and the fusion image in a form of a family tree. ### Claim 14 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the using of the randomly determined information includes applying fluctuation by using a random noise. ### Claim 15 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the using of the randomly determined information includes randomly determining an element of the fusion latent information. ### Claim 16 (depends on 1) The image generation apparatus as claimed in claim 1, wherein the use of the randomly determined information to generate the fusion latent information includes an application of fluctuation. ### Claim 17 An image generation method comprising: acquiring, by at least one processor, first latent information of a first image and second latent information of a second image, the first image including data related to a visual representation of a first object, and the second image including data related to a visual representation of a second object, generating, by the at least one processor, fusion latent information by using the first latent information and the second latent information, and generating, by the at least one processor, a fusion image by inputting the fusion latent information into a trained generative model, the fusion image including data related to a visual representation of a third object, wherein the generating of the fusion latent information includes generating, by the at least one processor, the fusion latent information by using randomly determined information, wherein the first object, the second object, and the third object are respectively a first character, a second character, and a third character, the first character, the second character, and the third character being different from each other, wherein the visual representation of the first object indicates an appearance of the first character, the visual representation of the second object indicates an appearance of the second character, and the visual representation of the third object indicates an appearance of the third character, and wherein the first latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the first character, the second latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the second character, and the fusion latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the third character. ### Claim 18 (depends on 17) The image generation method as claimed in claim 17, wherein the generating of the fusion image includes performing a genetic operation on the first latent information and the second latent information. ### Claim 19 (depends on 18) The image generation method as claimed in claim 18, wherein the genetic operation includes at least one of crossover, mutation, or selection. ### Claim 20 An image generation apparatus comprising: at least one memory; and at least one processor configured to: acquire first latent information of a first image and second latent information of a second image, the first image including data related to a visual representation of a first object, and the second image including data related to a visual representation of a second object, generate fusion latent information by using the first latent information and the second latent information, and generate a fusion image by inputting the fusion latent information into a trained generative model, the fusion image including data related to a visual representation of a third object, wherein the at least one processor displays, on a display device, the first image, the second image, and the fusion image in a form of a family tree, wherein the first object, the second object, and the third object are respectively a first character, a second character, and a third character, the first character, the second character, and the third character being different from each other, wherein the visual representation of the first object indicates an appearance of the first character, the visual representation of the second object indicates an appearance of the second character, and the visual representation of the third object indicates an appearance of the third character, and wherein the first latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the first character, the second latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the second character, and the fusion latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the third character. ### Claim 21 (depends on 20) The image generation apparatus as claimed in claim 20, wherein the at least one processor uses randomly determined information to generate the fusion latent information. ### Claim 22 (depends on 20) The image generation apparatus as claimed in claim 20, wherein the at least one processor displays, on the display device, the information of the at least one of the hair style, the hair color, the eye color, the skin color, the expression, or the attachment of the third character. ### Claim 23 An image generation method comprising: acquiring, by at least one processor, first latent information of a first image and second latent information of a second image, the first image including data related to a visual representation of a first object, and the second image including data related to a visual representation of a second object, generating, by the at least one processor, fusion latent information by using the first latent information and the second latent information, generating, by the at least one processor, a fusion image by inputting the fusion latent information into a trained generative model, the fusion image including data related to a visual representation of a third object, and displaying, by the at least one processor, on a display device, the first image, the second image, and the fusion image in a form of a family tree, wherein the first object, the second object, and the third object are respectively a first character, a second character, and a third character, the first character, the second character, and the third character being different from each other, wherein the visual representation of the first object indicates an appearance of the first character, the visual representation of the second object indicates an appearance of the second character, and the visual representation of the third object indicates an appearance of the third character, and wherein the first latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the first character, the second latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the second character, and the fusion latent information includes information of at least one of a hair style, a hair color, an eye color, a skin color, an expression, or an attachment of the third character. ### Claim 24 (depends on 23) The image generation method as claimed in claim 23, wherein the generating of the fusion latent information includes generating the fusion latent information by using randomly determined information. ### Claim 25 (depends on 23) The image generation method as claimed in claim 23, further comprising: displaying, by the at least one processor, on the display device, the information of the at least one of the hair style, the hair color, the eye color, the skin color, the expression, or the attachment of the third character.