Image generation system, image generation method, and program

Abstract

The goal is to obtain the desired image data appropriately and efficiently. [Solution] The image generation system 1 comprises an acquisition unit, a first generation control unit, a second generation control unit, and an evaluation unit. The acquisition unit acquires first concept information having the image concept desired by the user. The first generation control unit inputs the acquired first concept information into a first generation AI model that generates image data based on the concept information to generate multiple image data. The second generation control unit inputs the generated image data into a second generation AI model that generates concepts based on the image data to generate second concept information. The evaluation unit evaluates the degree of similarity between the first concept information and the second concept information for each image data.

Description

Technical Field

[0005] ,

[0001] The present invention relates to an image generation system, an image generation method, and a program.

Background Art

[0002] Conventionally, techniques for generating image data of a concept desired by a user are known. For example, an image selection device that provides a photograph along with the wishes of a photographer is known (see Patent Document 1). The image selection device captures a moving image of a subject from a camera at a predetermined shooting time, subdivides it at predetermined time intervals, and extracts a plurality of candidate images. The image selection device determines the direction of the face in the person image of each candidate image, and calculates an evaluation value for each candidate image based on the determination result. The image selection device selects an image with a desired face direction from among the plurality of candidate images based on the calculated evaluation value.

[0003] Also, a face image processing device that automatically determines the expression of a face and obtains a desired image is known (see Patent Document 2). The face image processing device detects a face image, inputs images of a plurality of persons including the face image, and determines whether the face of each of the plurality of persons is facing forward or the state of opening and closing of the pupils. The face image processing device selects and outputs an image with the highest evaluation value in the facial expression of each person.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] The conventional image selection device described above selects an image with a face orientation that meets the desired criteria. The conventional face image processing device described above selects an image where the face is facing forward and the pupils are open. Therefore, while it is possible to obtain an image that meets the desired face orientation, it is not possible to obtain an image that meets any arbitrary criteria. Furthermore, each of the conventional devices described above selects the desired image from the images that have actually been captured, and is not possible to generate a new image that meets the desired criteria.

[0006] In recent years, artificial intelligence (AI) for image generation has been developing rapidly. Artificial intelligence (AI) for image generation is a service or software that automatically generates finished image data simply by providing text or image data describing the desired image or atmosphere. However, it is currently quite difficult to get AI to generate an image that matches the user's intended concept. While some control over image generation is possible through text prompts, there are limitations to its accuracy. Furthermore, a characteristic of AI for image generation is that the output image data differs each time, even when using the same prompt. Therefore, it was necessary to repeatedly generate images until the intended image data was produced.

[0007] The objective of this invention is to obtain desired image data appropriately and efficiently. [Means for solving the problem]

[0008] To solve the above problem, the image generation system of the invention described in claim 1 is: An acquisition unit that acquires first concept information having the image concept desired by the user, A first generation control unit inputs the acquired first concept information into a first generation AI model that generates image data based on the concept information, and generates a predetermined number of image data. A second generation control unit inputs the generated image data into a second generation AI model that generates concepts based on the image data to generate second concept information, The system includes an evaluation unit that evaluates the degree of similarity between the first concept information and the second concept information for each of the aforementioned image data.

[0009] The invention described in claim 2 is an image generation system described in claim 1, The first generation AI model generates image data based on element information and concept information that have image elements to be included in the generated image data. The acquisition unit acquires element information and first concept information having the elements desired by the user. The first generation control unit inputs the acquired element information and the first concept information into the first generation AI model to generate a plurality of image data, The evaluation unit evaluates the degree of similarity between the element information obtained from each image data and the acquired element information.

[0010] The invention described in claim 3 is an image generation system described in claim 1, The system includes a display control unit that displays the aforementioned image data on a display unit.

[0011] The invention described in claim 4 is an image generation system described in claim 3, The display control unit displays either the image data with the highest evaluation result, or a plurality of image data with high evaluation results and the evaluation results corresponding to said image data.

[0012] The invention described in claim 5 is an image generation system described in claim 1, The evaluation unit uses a synonym database to generate an evaluation value for each image data, with the degree of similarity between the concepts of the first concept information and the second concept information being higher.

[0013] The invention described in claim 6 is an image generation system described in claim 1, The evaluation unit inputs the first concept information and the second concept information into the LLM for each of the image data, and obtains a higher evaluation value as the degree of similarity is higher.

[0014] The invention according to claim 7 is an image generation system according to claim 1, The predetermined number of sheets is the preset number of sheets or the number of sheets operationally input via an operation unit.

[0015] The invention according to claim 8 is an image generation system according to claim 1, It includes a first providing unit that provides the first generation AI model and the second generation AI model.

[0016] The invention according to claim 9 is an image generation system according to claim 6, It includes a second providing unit that provides the LLM.

[0017] The image generation method of the invention according to claim 10 is An acquisition unit that acquires first concept information having the concept of an image desired by a user, A first generation control step of inputting the acquired first concept information into a first generation AI model that generates image data based on the concept information to generate a plurality of image data, A second generation control step of inputting the generated image data into a second generation AI model that generates a concept based on the image data to generate second concept information, An evaluation step of evaluating the degree of similarity between the first concept information and the second concept information for each of the image data, is included.

[0018] The program of the invention according to claim 11 is An acquisition unit that acquires first concept information having the concept of an image desired by a user, A first generation control unit that inputs the acquired first concept information into a first generation AI model that generates image data based on the concept information to generate a predetermined number of image data, A second generation control unit that inputs the generated image data into a second generation AI model that generates a concept based on the image data to generate second concept information. An evaluation unit that evaluates the degree of similarity between the first concept information and the second concept information for each of the image data. Function as.

Effect of the Invention

[0019] According to the present invention, desired image data can be obtained appropriately and efficiently.

Brief Description of the Drawings

[0020] [Figure 1] It is a block diagram showing an image generation system according to a first embodiment of the present invention. [Figure 2] It is a block diagram showing the functional configuration of a server. [Figure 3] It is a block diagram showing the functional configuration of a terminal device. [Figure 4] It is a flowchart showing an image providing process. [Figure 5] It is a flowchart showing a concept evaluation process. [Figure 6] It is a diagram showing a display screen. [Figure 7] It is a diagram showing a first image, first concept information, second image, and second concept information of the first embodiment. [Figure 8] It is a diagram showing first concept information, second image, and second concept information of the second embodiment. [Figure 9] It is a diagram showing an example of information exchanged between an AI providing device and a server.

Mode for Carrying Out the Invention

[0021] The advantages and features provided by one or more embodiments of the present invention will be better understood from the following detailed description and accompanying drawings. However, these drawings are for illustrative purposes only and are not intended to define any limitations of the invention. Embodiments of the present invention will be described below with reference to the drawings. However, the scope of the invention is not limited to the disclosed embodiments.

[0022] (First Embodiment) A first embodiment of the present invention will be described with reference to Figures 1 to 8. First, the device configuration of the image generation system 1 of this embodiment will be described with reference to Figures 1 to 3. Figure 1 is a block diagram of the image generation system 1 of this embodiment. Figure 2 is a block diagram of the functional configuration of the server 10. Figure 3 is a block diagram of the functional configuration of the terminal device 20.

[0023] As shown in Figure 1, the image generation system 1 is a system that generates image data of a concept intended by the user. The image generation system 1 comprises a server 10, a terminal device 20, and an AI providing device 30. The AI ​​providing device 30 functions as a first providing unit and a second providing unit. The server 10, the terminal device 20, and the AI ​​providing device 30 are connected to each other so as to be able to communicate via a communication network 40. The communication network 40 is, for example, the Internet, but is not limited to this. The communication network 40 may be a wired communication network, a wireless communication network, or a LAN (Local Area Network).

[0024] Server 10 is an information processing device that provides image data of the user's intended concept to terminal device 20. Terminal device 20 is a desktop PC (Personal Computer) used by the user. Terminal device 20 is not limited to a desktop PC, but may be other information processing devices such as a palmtop PC or a smartphone.

[0025] The AI ​​providing device 30 is an information processing device that provides a first generation AI model and a second generation AI model to external devices such as a server 10. The first generation AI model is a trained model of a so-called image generation AI. When a first prompt is input, the first generation AI model automatically generates image data. The generated image data is image data that closely resembles the first concept information of the input first prompt. A prompt is natural language text that describes the task that the generation AI should perform. The first prompt includes first concept information that mainly indicates the finished image or atmosphere intended by the user. The first concept information includes at least one concept in the finished image intended by the user. The first prompt has a description of a task that causes the AI ​​to generate image data that satisfies the first concept information (and item information of items included in the finished image).

[0026] The second generative AI model is a trained model that generates second conceptual information in text form when given image data and a second prompt as input. The second conceptual information contains concepts analyzed from the image data. The second prompt contains text for a standardized task that prompts the second generative AI model to generate the second conceptual information.

[0027] The AI ​​provider 30 receives, for example, a first prompt from the server 10. The AI ​​provider 30 inputs the first prompt into the first generation AI model and generates image data that closely matches the first concept information. The AI ​​provider 30 sends the generated image data to the server 10, the source of the first prompt. When the AI ​​provider 30 receives the image data from the server 10, it inputs the image data into the second generation AI model and generates second concept information that satisfies the image data. The AI ​​provider 30 sends the generated second concept information to the server 10, the source of the image data.

[0028] Next, with reference to Figure 2, the internal functional configuration of the server 10 will be explained. As shown in Figure 2, the server 10 has a control unit 11, an operation unit 12, a storage unit 13, a display unit 14, and a communication unit 15. Each part of the server 10 is connected to the others via a bus. The control unit 11 functions as an acquisition unit, a first generation control unit, a second generation control unit, an evaluation unit, and a display control unit.

[0029] The control unit 11 controls various parts of the server 10. The control unit 11 has a CPU (Central Processing Unit) and RAM (Random Access Memory). The control unit 11 reads various programs stored in the memory unit 13, loads them into RAM, and executes various processes in cooperation with the CPU and the loaded programs.

[0030] The operation unit 12 includes a keyboard and a pointing device such as a mouse. The operation unit 12 receives key input and position input from the user and outputs the operation information to the control unit 11.

[0031] The storage unit 13 is an HDD (Hard Disk Drive), SSD (Solid State Drive), etc. The storage unit 13 stores information such as data in a read-and-write manner. In particular, the storage unit 13 stores an image provision program. The image provision program is a program for executing the image provision process described later. The storage unit 13 also stores a thesaurus database. The thesaurus database is data from a thesaurus and contains synonyms that have a similar meaning to any given term. The thesaurus database is used to evaluate the similarity between the concept of the first concept information and the concept of the second concept information. Furthermore, the thesaurus database may also contain information on the degree of similarity (degree) between the term and its synonyms. In this configuration, the evaluation of similarity can be subdivided.

[0032] The display unit 14 has a display panel such as an LCD (liquid crystal display) or an ELD (electro-luminescent display). The display unit 14 displays display information on the display panel in accordance with the control of the control unit 11.

[0033] The communication unit 15 is a communication module such as a network card, and performs wired or wireless communication with external devices such as terminal devices 20 and AI provisioning devices 30 on the communication network 40. The control unit 11 transmits and receives information with the terminal devices 20 and AI provisioning devices 30 via the communication unit 15.

[0034] Next, with reference to Figure 3, the internal functional configuration of the terminal device 20 will be explained. As shown in Figure 3, the terminal device 20 has a control unit 21, an operation unit 22, a storage unit 23, a display unit 24, and a communication unit 25. Each part of the terminal device 20 is connected to the others via a bus.

[0035] The control unit 21 controls various parts of the terminal device 20. The control unit 21 has a CPU and RAM. The control unit 21 reads various programs stored in the memory unit 23, loads them into RAM, and executes various processes in cooperation with the CPU and the loaded programs.

[0036] The operation unit 22 includes a keyboard and a pointing device such as a mouse. The operation unit 22 receives key input and position input from the user and outputs the operation information to the control unit 21.

[0037] The storage unit 23 is an HDD, SSD, etc. The storage unit 23 stores various types of information in a read-and-write manner.

[0038] The display unit 24 has a display panel such as an LCD or ELD. The display unit 24 displays display information on the display panel in accordance with the control of the control unit 21.

[0039] The communication unit 25 is a communication module such as a network card, and performs wired or wireless communication with external devices such as the server 10 on the communication network 40. The control unit 21 sends and receives information with the server 10 and other devices via the communication unit 25.

[0040] Next, the operation of the image generation system 1 will be explained with reference to Figures 4 to 8. Figure 4 is a flowchart of the image provision process. Figure 5 is a flowchart of the concept evaluation process. Figure 6 shows the display screens 50 and 60. Figure 7 shows the first image 71, first concept information 72, second images 73, 75, 77, 79, and second concept information 74, 76, 78, 80 of the first embodiment. Figure 8 shows the first concept information 82, second images 83, 85, and second concept information 84, 86 of the second embodiment.

[0041] Referring to Figure 4, the image provision process performed on server 10 will be explained. The image provision process is the process of generating image data that closely matches the user's intended concept, or the concept and item information, and providing it to terminal device 20. First, in terminal device 20, control unit 21 receives input from the user via operation unit 22, either first concept information or first concept information and item information. The first concept information is text containing the image concept of the image data to be generated. The item information is text of the items as image elements that the user wishes to include in the final image. For example, if the user wishes to receive image data of an advertisement, the first concept information would include the concept of the final image and atmosphere that the user wants to include in the advertisement image. However, the item information may also be information on other image elements, such as scenery other than the items to be included in the final image.

[0042] The control unit 21 transmits the input first concept information, or the first concept information and item information, to the server 10 via the communication unit 25. The control unit 11 of the server 10 begins receiving the input first concept information, or the first concept information and item information, from the terminal device 20 via the communication unit 15. Triggered by this start of reception, the control unit 11 executes image provision processing according to the image provision program stored in the storage unit 13.

[0043] First, the control unit 11 determines whether or not the reception of item information has started (step S11). If the reception of item information has started (step S11; YES), the control unit 11 completes the reception of the first concept information and the item information (step S12). The control unit 11 generates a first prompt including the first concept information and the item information received in step S12 (step S13). The first concept information and the item information are stored, for example, in the storage unit 13.

[0044] If item information has not been started to be entered (step S11; NO), the control unit 11 completes the reception of the first concept information (step S14). The control unit 11 generates a first prompt containing the first concept information received in step S14 (step S15). The control unit 11 transmits the first prompt to the AI ​​providing device 30 for the first generated AI model via the communication unit 15 (step S16). The AI ​​providing device 30 receives the first prompt in step S13 or S15 from the server 10. The AI ​​providing device 30 inputs the first prompt in step S13 or step S15 into the first generated AI model to generate one image data. The AI ​​providing device 30 transmits the generated image data to the server 10. In step S16, the control unit 11 receives the generated image data from the AI ​​providing device 30 via the communication unit 15.

[0045] The control unit 11 determines whether a predetermined number of image data images have been generated from the start of the image generation process in step S16 (step S17). The predetermined number is a preset number of image data images to be generated in the image provision process. The predetermined number is stored in the storage unit 13, for example. The predetermined number may also be included in the first prompt of step S15. If the predetermined number of images have not been generated (step S17; NO), the control unit 11 generates a second prompt (step S18). The second prompt is a prompt for the second generated AI model. The second prompt is stored in the storage unit 13, for example, and is a standard prompt for obtaining second concept information from the image data. In step S18, the image data and the second prompt are transmitted to the AI ​​provision device 30 for the second generated AI model via the communication unit 15.

[0046] The AI ​​provider 30 receives the image data and second prompt from the server 10 in step S18. The AI ​​provider 30 inputs the image data and second prompt from step S18 into the second generation AI model to generate second concept information. The second concept information is the text of the concept representing the image of the image data. The AI ​​provider 30 transmits the generated second concept information to the server 10. The control unit 11 receives the generated second concept information from the AI ​​provider 30 via the communication unit 15 (step S19).

[0047] The control unit 11 determines whether or not there is item information received in step S12 (step S20). If element information is available (step S20; YES), the control unit 11 performs image analysis on the image data received in step S16 to obtain item information of the included items (step S21). In step S21, the control unit 11 performs an item evaluation by comparing the analyzed item information with the input item information in step S12 and evaluating the degree of similarity. The item evaluation calculates an evaluation value, for example, where the higher the degree of similarity of the item information, the higher the numerical value. The second generating AI may also have a function to perform image analysis on items included in the input image data and output item information of those items. In this configuration, the second prompt in step S18 includes text requesting item information from the second image data. In step S19, the item information from the second image data is received from the AI ​​providing device 30 along with the second concept information. In step S21, the item information from step S19 is compared with the item information from step S12 to generate the evaluation result of the item evaluation. For product evaluation, a synonym database may be used, or Large Language Models (LLMs), as described later, may be employed.

[0048] The control unit 11 performs a concept evaluation process to evaluate the second concept information (step S22). The concept evaluation process in step S22 will now be explained with reference to Figure 5. The control unit 11 compares the concepts included in the first concept information with the concepts included in the second concept information (step S31). In step S31, the control unit 11 determines whether or not there is a concept that perfectly matches based on the comparison result. If there is a matching concept (step S31; YES), the control unit 11 sets an evaluation value A as the evaluation result of the concept evaluation (step S32). The concept evaluation process ends. Evaluation value A and evaluation values ​​B, C, D, and E, which will be described later, are, for example, numerical values. Here, we assume evaluation value A > evaluation value B > evaluation value C > evaluation value D > evaluation value E.

[0049] If no matching concept is found (step S31; NO), the control unit 11 refers to the synonym database stored in the memory unit 13 (step S33). In step S33, the control unit 11 compares the concept included in the first concept information with the synonyms of the concept included in the second concept information. In step S33, the control unit 11 compares the synonyms of the concept included in the first concept information with the concept included in the second concept information. Alternatively, in step S33, the control unit 11 may compare the synonyms of the concept included in the first concept information with the synonyms of the concept included in the second concept information. In the synonym database, each synonym is associated with a synonym level L1 to L3, which indicates the degree of similarity. In step S33, the control unit 11 determines whether or not there is a matching concept based on the comparison result.

[0050] If a matching concept exists (step S33; YES), the control unit 11 refers to the synonym levels L1 to L3 of the synonyms of the matching concept (step S34). In step S34, the control unit 11 sets evaluation values ​​B to C as evaluation results of the concept evaluation based on the referenced synonym levels L1 to L3. The concept evaluation process ends. For example, evaluation value B is set corresponding to synonym level L1. Evaluation value C is set corresponding to synonym level L2. Evaluation value D is set corresponding to synonym level L3.

[0051] If no matching concept is found (step S33; NO), the control unit 11 sets an evaluation value E as the evaluation result of the concept evaluation (step S35). The concept evaluation process ends. Note that the evaluation value is not limited to five levels, but may be any other multiple levels or a numerical value itself.

[0052] Returning to Figure 4, if there is no element information (step S20; NO), the process proceeds to step S22. The control unit 11 associates the image data from step S16 with the evaluation results of the item evaluation and concept evaluation from steps S21 and S22 and stores them in the storage unit 13 (step S23). The evaluation results of the item evaluation and concept evaluation are combined into a single evaluation value, for example, as the sum of the evaluation values ​​for the item evaluation and concept evaluation. The information associated with the image data includes the first concept information from step S12, the first prompt from step S13 or S15, the second prompt from step S18, and the second concept information from step S19.

[0053] The control unit 11 determines whether the final flag, which indicates that a predetermined number of image data have been generated, is on (step S24). If the final flag is off (step S24; NO), the process proceeds to step S16. If the predetermined number of images have been generated (step S17; YES), the control unit 11 turns on the final flag (step S25). The process proceeds to step S18.

[0054] If the final flag is on (step S24; YES), the control unit 11 reads the display image data, etc. from the storage unit 13 (step S26). In step S26, the control unit 11 transmits the read display image data, etc. to the terminal device 20 via the communication unit 15. The control unit 21 of the terminal device 20 receives the display image data, etc. from the server 10 via the communication unit 25 and displays it on the display unit 14. The image provision process is completed.

[0055] Here, referring to Figure 6, the display screens 50 and 60 that are displayed in accordance with step S26 will be explained. First, if only the image data with the highest evaluation value is to be displayed on the terminal device 20, for example, display screen 50 will be displayed on the terminal device 20. In this case, in step S26, the control unit 11 reads and refers to all evaluation results and extracts the highest evaluation value. The control unit 11 reads the image data with the highest evaluation value and its first concept information from the storage unit 13 and transmits it to the terminal device 20. The control unit 21 of the terminal device 20 receives the image data with the highest evaluation value and the first concept information from the server 10, generates display screen 50, and displays it on the display unit 14.

[0056] Display screen 50 is the display screen for the image data with the highest evaluation value in the evaluation results. Display screen 50 has first concept information 51 and an image 52. The first concept information 51 includes, for example, three concepts "aaa", "bbb", and "ccc". Image 52 is the image of the image data with the highest evaluation value among a predetermined number of image data generated in correspondence with the first concept information 51.

[0057] Next, when displaying the top predetermined number of image data with the highest evaluation values ​​on the terminal device 20, for example, a display screen 60 is displayed on the terminal device 20. In this case, in step S26, the control unit 11 reads and references all evaluation results and extracts the top predetermined number of evaluation values. The control unit 11 reads the top predetermined number of image data, their first concept information, evaluation values, and second concept information from the storage unit 13 and transmits them to the terminal device 20. The control unit 21 of the terminal device 20 receives the top predetermined number of image data, their first concept information, evaluation values, and second concept information from the server 10. The control unit 21 generates a display screen 60 and displays it on the display unit 14.

[0058] Display screen 60 is a display screen for image data that is the top predetermined number (=2) in evaluation value. However, the top predetermined number is not limited to 2. Display screen 60 has first concept information 61, image 62, image information 63, image 64, and image information 65. First concept information 61 includes, for example, three concepts "aaa", "bbb", and "ccc". Image 62 is the image of the image data with the highest evaluation value among a predetermined number of image data generated in correspondence with the first concept information 61. Image information 63 is text corresponding to the image data of image 62. Image information 63 includes the title "First Candidate" which includes the ranking (1st place) of the evaluation value, the evaluation value, and second concept information. The second concept information of image information 63 includes, for example, two concepts "aaa" and "bbb".

[0059] Image 64 is the image of the image data with the second highest evaluation value among a predetermined number of image data generated in accordance with the first concept information 61. Image information 65 is text information corresponding to the image data of image 62. Image information 63 includes the title "Second Candidate" including the ranking of the evaluation value (2nd place), the evaluation value, and the second concept information. The second concept information of image information 65 includes, for example, three concepts "ddd", "aaa", and "bbb". The display screen 50 (display screen 60) may include at least one of the following: the first concept information, the second concept information, the evaluation value, the item information, the first prompt, and the second prompt, corresponding to image 52 (images 62, 64).

[0060] Next, with reference to Figure 7, a first embodiment of this model will be described. The first embodiment is an example in which image data is generated using item information and first concept information in the image generation process.

[0061] User 70 has already created the image for the intended first image 71 and sends it to the server 10. The first image 71 is a poster image for a cafe menu and includes items such as a beer mug and a coffee cup that the user wants to include in the completed second image. Furthermore, in accordance with step S12, the terminal device 20 receives the first concept information 72 and item information intended by user 70 and sends them to the server 10.

[0062] The loop from steps S16 to S24 generates image data for the second images 73, 75, 77, and 79, corresponding to each step S16. Similarly, the second concept information 74, 76, 78, and 80 are generated, corresponding to each step S19. The second concept information 74 corresponds to the second image 73. The second concept information 76 corresponds to the second image 75. The second concept information 78 corresponds to the second image 77. The second concept information 80 corresponds to the second image 79.

[0063] Second image 73 includes the items "beer mug" and "coffee cup" that are included in first image 71, and the evaluation value for the item evaluation in step S21 is high. Second concept information 74 includes multiple concepts. To the left of each concept, the evaluation values ​​A to C for the concept evaluation in step S22 corresponding to that concept are displayed. The evaluation values ​​here are on a three-point scale. Second images 75, 77, 79 and second concept information 76, 78, 80 are the same as second image 73 and second concept information 74.

[0064] In the example in Figure 7, the user already has a certain image in mind. For example, even in situations where a detailed image needs to be provided and the user doesn't have time to create the image, image data and evaluation results can still be generated.

[0065] Referring to Figure 8, a second embodiment of this model will be described. The second embodiment is an example in which image data is generated using only the first concept information in the image generation process. User 81 has not decided on the desired image image, but has decided on the purpose and concept.

[0066] In response to step S12 of the image generation process, first concept information 82 corresponding to the user 81's intended purpose and concept is input to the terminal device 20 and transmitted to the server 10.

[0067] The loop from steps S16 to S24 generates image data for the second images 83 and 85, corresponding to each step S16. Similarly, second concept information 84 and 86 are generated, corresponding to each step S19. Second concept information 84 corresponds to the second image 83. Second concept information 86 corresponds to the second image 85.

[0068] The second concept information 84 includes multiple concepts. To the left of each concept, the evaluation values ​​A to C for the concept evaluation in step S22 corresponding to that concept are displayed. The second image 85 and the second concept information 86 are the same as the second image 83 and the second concept information 84.

[0069] In the example in Figure 8, the user does not have an image of the picture and must create the image data. Even in this situation, the image data and evaluation results can be generated. This prevents the process from taking too long, as the user lacks the ability to create the image of the provided picture themselves.

[0070] As described above, according to this embodiment, the server 10 of the image generation system 1 includes a control unit 11. The control unit 11 acquires first concept information having an image concept desired by the user. The control unit 11 inputs a first prompt containing the acquired first concept information to a first generation AI model that generates image data based on the concept information, and generates multiple image data. The control unit 11 inputs the generated image data to a second AI model that generates concepts based on the image data, and generates second concept information. The control unit 11 evaluates the degree of similarity between the first concept information and the second concept information for each image data. Therefore, prompts can be easily generated, and image data of the concept desired by the user can be obtained appropriately and efficiently according to the evaluation result.

[0071] The first generation AI model generates image data based on element information (item information) and concept information that include image elements (items) to be included in the generated image data. The control unit 11 acquires the item information and first concept information desired by the user. The control unit 11 inputs the acquired item information and first concept information into the first generation AI model to generate multiple image data. The control unit 11 evaluates the degree of similarity between the item information obtained from each image data and the acquired element information. As a result, it is possible to obtain the desired image data that reflects the item information and first concept information, and to evaluate whether the item information is reflected in each image data.

[0072] The control unit 11 of the server 10 transmits image data to the terminal device 20 and displays it on the display unit 24. This allows the user to visually confirm the image data.

[0073] The control unit 11 displays the image data with the highest evaluation result. Alternatively, the control unit 11 displays multiple image data with high evaluation results and the evaluation results corresponding to each of those image data. This allows the user to easily identify the image data with the highest evaluation result, or to identify multiple image data with high evaluation results along with their respective evaluation results.

[0074] The control unit 11 uses a synonym database to generate an evaluation value for each image data, with the higher the degree of similarity between the first concept information and the second concept information, the higher the evaluation value. Therefore, the degree to which the second concept information of the image data is similar to the first concept information can be accurately evaluated.

[0075] The predetermined number of image data files is set in advance. This reduces the burden on the user.

[0076] The image generation system 1 includes an AI providing device 30 that provides a first generation AI model and a second generation AI model. This simplifies the configuration of the server 10.

[0077] (Second Embodiment) A second embodiment of the present invention will be described with reference to Figure 9. Figure 9 is a diagram showing an example of information exchanged between the AI ​​providing device 30 and the server 10.

[0078] In the first embodiment described above, the server 10 stored a synonym database and performed concept evaluation processing using the synonym database. In this embodiment, the AI ​​providing device 30 provides the LLM, and the server 10 performs concept evaluation processing using the LLM.

[0079] The device configuration of this embodiment is the same as in the first embodiment, using the image generation system 1. However, the AI ​​providing device 30 provides the LLM. The LLM of the AI ​​providing device 30 is an AI into which the first concept information is input while the second concept information has been generated. The LLM performs a concept evaluation by comparing the second concept information with the first concept information using synonyms, similar to step S22 of the image provisioning process, and outputs the evaluation result. For this reason, the server 10 does not store the synonym database in the storage unit 13.

[0080] Next, the operation of the image generation system 1 will be described with reference to Figure 9. Similar to the first embodiment, the image provision process shown in Figures 4 and 5 is performed by the server 10. Here, we will mainly describe the parts that differ from the image provision process of the first embodiment, and omit the explanation of the parts that are the same.

[0081] Steps S11 to S21 of the image provision processing are the same as in the first embodiment. In the concept evaluation processing of step S22, the control unit 11 of the server 10 acquires the first concept information of step S12 and the second concept information of step S19. The control unit 11 transmits the acquired first concept information and second concept information to the AI ​​provisioning device 30 for LLM via the communication unit 15.

[0082] The AI ​​providing device 30 receives the first concept information and the second concept information from the server 10 and inputs them into the LLM. The LLM compares the received second concept information with the received first concept information, calculates an evaluation value as the result of the concept evaluation, and transmits it to the server 10. Steps S23 to S26 of the image provisioning process are the same as in the first embodiment.

[0083] Here, with reference to Figure 9, an embodiment of this design will be described. In this case, corresponding to step S11 of the image provision process, the user inputs "refreshing feeling," "freedom feeling," and "coolness" as the concepts for the first concept information. The user wishes to generate a second image for a poster. Also, corresponding to step S11, the item information "beer mug" has been input.

[0084] In step S16, image data of the second image 111 is generated. In step S18, the control unit 11 of the server 10 generates a second prompt 110 and sends the second prompt 110 along with the image data of the second image 111 to the AI ​​providing device 30. Here, the second prompt 110 includes text requesting the image content of the second image 111, along with text requesting the second concept information. The second generating AI model of the AI ​​providing device 30 generates item information 121 and second concept information 122 for the second image 111 from the received image data of the second image 111 and the second prompt 110. The second concept information 120 includes the concepts "refreshment," "coolness," "fun," and "liberation," along with text describing the image content of the second image 111. In step S21, the control unit 11 compares the item information "beer mug" received in step S12 with the item information 121 and generates an evaluation result for the item evaluation.

[0085] In step S22, the control unit 11 generates a prompt 130 and sends it to the LLM of the AI ​​providing device 30. The prompt 130 includes first concept information and a request for a three-stage evaluation method in concept evaluation. The LLM obtains four concepts of second concept information 120 from the second image generation AI model or server 10. Using the received prompt 130, the LLM generates evaluation results 140 of the second concept information. The LLM sends the evaluation results 140 to the server 10. In response to step S26, the image data with the highest total score of, for example, the evaluation results of the item evaluation and the evaluation results 140 from a predetermined number of generated image data is displayed on the terminal device 20.

[0086] As described above, according to this embodiment, the control unit 11 inputs the first concept information and the second concept information for each image data into the LLM, and obtains a higher evaluation value the higher the degree of similarity. Therefore, the degree to which the second concept information of the image data is similar to the first concept information can be evaluated more accurately.

[0087] Furthermore, the image generation system 1 includes an AI providing device 30 that provides LLM. This simplifies the configuration of the server 10.

[0088] The above description discloses an example in which a storage unit (HDD, SSD) is used as a computer-readable medium for the program according to the present invention, but the invention is not limited to this example. Other computer-readable mediums that can be used include non-volatile memory such as flash memory and portable recording media such as CD-ROMs. Furthermore, carrier waves can also be used as a medium for providing program data according to the present invention via a communication line.

[0089] The above-described embodiments are merely examples of the image generation system, image generation method, and program according to the present invention, and are not limited thereto. For example, the first and second embodiments may be combined as appropriate.

[0090] Furthermore, in the above embodiment, the AI ​​providing device 30 is configured to provide the first generated AI, the second generated AI, and LLM, but the configuration is not limited to this. The server 10 itself may be configured to have the function of providing the first generated AI, the second generated AI, and LLM.

[0091] Furthermore, in the above embodiment, the number of image data generated was a predetermined number, but it is not limited to this. A characteristic of image generation AI is that even with the same prompt, the generated image data will be different each time. Therefore, it is reasonable to increase the number of trials, i.e., the number of image data generated (predetermined number), in order to obtain the desired image data. The control unit 21 of the terminal device 20 may accept input from the user to set a predetermined number of image data via the operation unit 22. For example, when the control unit 21 accepts input of first concept information from the user via the operation unit 22, it may also accept input of first concept information including the number of image data to be generated. The number of image data to be generated input by the terminal device 20 is transmitted to the server 10. The control unit 11 of the server 10 sets the predetermined number of received image data to the predetermined number in step S17 of the image provision process. In other words, the predetermined number of image data is the number of data entered by the user via the operation unit 22. Therefore, the user can specify the number of image data to be generated. By increasing the predetermined number, the completeness of the image data for the desired concept can be improved. Alternatively, the number of images can be reduced to shorten the time required for image data generation and evaluation.

[0092] While embodiments of the present invention have been described and illustrated in detail, the disclosed embodiments are for illustrative and illustrative purposes only and are not limiting. The scope of the present invention should be interpreted by the terms of the appended claims. [Explanation of Symbols]

[0093] 1. Image generation system 10 servers 11 Control Unit 12 Control section 13 Storage section 14 Display section 15 Communications Department 20 Terminal devices 21 Control Unit 22 Control section 23 Memory section 24 Display section 25 Communications Department 30 AI providing device 40 Communication Networks

Claims

1. An acquisition unit that acquires first concept information having the image concept desired by the user, A first generation control unit inputs the acquired first concept information into a first generation AI model that generates image data based on the concept information, and generates a predetermined number of image data. A second generation control unit inputs the generated image data into a second generation AI model that generates concepts based on the image data to generate second concept information, An image generation system comprising: an evaluation unit that evaluates the degree of similarity between the first concept information and the second concept information for each of the aforementioned image data.

2. The first generation AI model generates image data based on element information and concept information that have image elements to be included in the image data to be generated. The acquisition unit acquires element information and first concept information having the elements desired by the user. The first generation control unit inputs the acquired element information and the first concept information into the first generation AI model to generate a plurality of image data. The image generation system according to claim 1, wherein the evaluation unit evaluates the degree of similarity between element information obtained from each of the image data and the acquired element information.

3. The image generation system according to claim 1, further comprising a display control unit that displays the image data on a display unit.

4. The image generation system according to claim 3, wherein the display control unit displays either the image data with the highest evaluation result, or a plurality of image data with high evaluation results and the evaluation results corresponding to said image data.

5. The image generation system according to claim 1, wherein the evaluation unit uses a synonym database to generate an evaluation value for each image data such that the degree of similarity in the concepts of the first concept information and the second concept information is higher.

6. The image generation system according to claim 1, wherein the evaluation unit inputs the first concept information and the second concept information to the LLM for each image data, and obtains a higher evaluation value the higher the degree of similarity.

7. The image generation system according to claim 1, wherein the predetermined number of images is the number of images set in advance or the number of images input via the operation unit.

8. The image generation system according to claim 1, further comprising a first providing unit that provides the first generation AI model and the second generation AI model.

9. The image generation system according to claim 6, further comprising a second providing unit that provides the aforementioned LLM.

10. An acquisition unit that acquires first concept information having the image concept desired by the user, The first generation control step involves inputting the acquired first concept information into a first generation AI model that generates image data based on the concept information to generate multiple image data files, and A second generation control step involves inputting the generated image data into a second generation AI model that generates concepts based on the image data to generate second concept information. An image generation method comprising an evaluation step of evaluating the degree of similarity between the first concept information and the second concept information for each of the aforementioned image data.

11. Computers An acquisition unit that acquires first concept information having the image concept desired by the user, The acquired first concept information is input to a first generation AI model that generates image data based on the concept information, and a first generation control unit generates a predetermined number of image data. A second generation control unit inputs the generated image data into a second generation AI model that generates concepts based on the image data to generate second concept information. An evaluation unit evaluates the degree of similarity between the first concept information and the second concept information for each of the aforementioned image data. A program designed to function as such.

Images