Synthetic data generation system and synthetic data generation program

Abstract

This invention provides a synthetic data generation system and program capable of presenting users with synthetic data that is effective for idea generation. [Solution] A synthetic data generation system according to one embodiment of the present disclosure includes a second learning model that, upon input of synthetic data obtained by inputting data to a first learning model and a judgment criterion, outputs a judgment result obtained by judging the synthetic data against the judgment criterion, and a processing unit that determines whether the judgment result satisfies a predetermined selection criterion and performs data processing to display a list of multiple synthetic data that satisfy the selection criterion.

Description

Technical Field

[0001] The present disclosure relates to a synthetic data generation system and a synthetic data generation program.

Background Art

[0002] Automobile designers seek unique design ideas from AI (artificial intelligence) while meeting the feasibility of the items they want to design. For example, Patent Document 1 discloses an AI capable of generating unique designs by multiplying an image of an item a user wants to design by an image of a natural object.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

[0004] The synthetic data generation system according to the first aspect of the present disclosure includes a first learning model, a second learning model, and a processing unit. When one or more first data and second data are input to the first learning model, the first learning model can output synthetic data obtained by multiplying the input one or more first data or one or more third data corresponding to the one or more first data by the input second data or fourth data corresponding to the second data. When the synthetic data obtained from the first learning model and a determination criterion are input to the second learning model, the second learning model can output a determination result obtained by determining the input synthetic data according to the input determination criterion. The processing unit can determine whether the determination result meets a predetermined selection criterion and perform data processing for list-displaying a plurality of synthetic data that meet the selection criterion.

[0005] The synthetic data generation program relating to the second aspect of this disclosure can cause a computer to execute the following four actions: (1) Accepts input of one or more first data and second data. (2) Inputting the received one or more first data and second data into the first learning model, thereby obtaining composite data from the first learning model that is a combination of one or more first data, or one or more third data corresponding to one or more first data, and the second data, or the fourth data corresponding to the second data. (3) By inputting the synthetic data obtained from the first learning model and the judgment criteria into the second learning model, the judgment result obtained by judging the synthetic data according to the judgment criteria is acquired from the second learning model. (4) Determine whether the judgment result meets the predetermined selection criteria, and perform data processing to display a list of multiple composite data that meet the selection criteria. [Brief explanation of the drawing]

[0006] The accompanying drawings are provided for further understanding of this disclosure and are incorporated herein and constitute part of this specification. The drawings illustrate one embodiment and, together with the specification, serve to illustrate the principles of this disclosure.

[0007] [Figure 1] Figure 1 is a diagram showing an example of a functional block of a synthetic data generation system according to the first embodiment of this disclosure. [Figure 2] Figure 2 shows an example of a learning model included in the generation AI section of Figure 1. [Figure 3] Figure 3 shows an example of a learning model included in the generation AI section of Figure 1. [Figure 4] Figure 4 shows an example of a learning model included in the generation AI section of Figure 1. [Figure 5] Figure 5 shows an example of the reference data in Figure 1. [Figure 6]Figure 6 shows an example of a learning model included in the filter section of Figure 1. [Figure 7] Figure 7 shows an example of the interface of the list display section in Figure 1. [Figure 8] Figure 8 shows an example of the interface of the list display section in Figure 1. [Figure 9] Figure 9 shows a modified example of the functional block of the synthetic data generation system shown in Figure 1. [Figure 10] Figure 10 shows an example of a learning model included in the generation AI section of Figure 9. [Figure 11] Figure 11 shows an example of a learning model included in the generative AI section of Figure 9. [Figure 12] Figure 12 shows an example of a learning model included in the generation AI section of Figure 9. [Figure 13] Figure 13 shows a modified example of the functional block of the synthetic data generation system shown in Figure 1. [Figure 14] Figure 14 is a diagram illustrating an example of data conversion in the data conversion unit of Figure 13. [Figure 15] Figure 15 is a diagram illustrating an example of data conversion in the data conversion unit of Figure 13. [Figure 16] Figure 16 is a diagram illustrating an example of data conversion in the data conversion unit of Figure 13. [Figure 17] Figure 17 shows a modified example of the functional block of the synthetic data generation system shown in Figure 1. [Figure 18] Figure 18 shows a modified example of the functional block of the synthetic data generation system shown in Figure 1. [Figure 19] Figure 19 is a diagram showing an example of the schematic configuration of the synthetic data generation system shown in Figure 1. [Figure 20] Figure 20 shows an example of the list display procedure in the synthetic data generation system shown in Figure 19. [Figure 21] Figure 21 shows a modified example of the schematic configuration of the synthetic data generation system shown in Figure 13. [Figure 22]FIG. 22 is a diagram showing an example of a list display procedure in the synthetic data generation system of FIG. 21. [Figure 23] FIG. 23 is a diagram showing a modified example of the schematic configuration of the synthetic data generation system of FIG. 1. [Figure 24] FIG. 24 is a diagram showing a modified example of the schematic configuration of the synthetic data generation system of FIG. 13. [Figure 25] FIG. 25 is a diagram showing an example of a functional block of the synthetic data generation system according to the second embodiment of the present disclosure. [Figure 26] FIG. 26 is a diagram showing an example of an interface of the selection flag registration unit of FIG. 25. [Figure 27] FIG. 27 is a diagram showing an example of an interface of the selection flag registration unit of FIG. 25. [Figure 28] FIG. 28 is a diagram showing an example of the reference data of FIG. 25. [Figure 29] FIG. 29 is a diagram showing an example of a learning model included in the filter unit of FIG. 25. [Figure 30] FIG. 30 is a diagram showing an example of an interface of the list display unit of FIG. 25. [Figure 31] FIG. 31 is a diagram showing an example of an interface of the list display unit of FIG. 25. [Figure 32] FIG. 32 is a diagram showing a modified example of the functional block of the synthetic data generation system of FIG. 25. [Figure 33] FIG. 33 is a diagram showing a modified example of the functional block of the synthetic data generation system of FIG. 25. [Figure 34] FIG. 34 is a diagram showing an example of the schematic configuration of the synthetic data generation system of FIG. 25. [Figure 35] FIG. 35 is a diagram showing an example of a list display procedure in the synthetic data generation system of FIG. 34. [Figure 36] FIG. 36 is a diagram showing a modified example of the schematic configuration of the synthetic data generation system of FIG. 33. [Figure 37] FIG. 37 is a diagram showing an example of a list display procedure in the synthetic data generation system of FIG. 36. [Figure 38] Figure 38 shows a modified example of the schematic configuration of the synthetic data generation system shown in Figure 25. [Figure 39] Figure 39 shows a modified example of the schematic configuration of the synthetic data generation system shown in Figure 33. [Figure 40] Figure 40 is a diagram showing an example of a functional block of a synthetic data generation system according to a third embodiment of the present disclosure. [Figure 41] Figure 41 shows a modified example of the functional block of the synthetic data generation system shown in Figure 40. [Figure 42] Figure 42 is a diagram showing an example of the schematic configuration of the synthetic data generation system shown in Figure 40. [Figure 43] Figure 43 is a diagram illustrating an example of the list display procedure in the synthetic data generation system shown in Figure 42. [Figure 44] Figure 44 is a diagram showing an example of the schematic configuration of the synthetic data generation system shown in Figure 41. [Figure 45] Figure 45 is a diagram illustrating an example of the list display procedure in the synthetic data generation system shown in Figure 44. [Figure 46] Figure 46 shows a modified example of the schematic configuration of the synthetic data generation system shown in Figure 40. [Figure 47] Figure 47 shows a modified example of the schematic configuration of the synthetic data generation system shown in Figure 41. [Modes for carrying out the invention]

[0008] <1. Background> Automotive designers are looking to AI to generate unconventional design ideas while also ensuring the feasibility of the items they want to design. For example, the aforementioned Patent Document 1 discloses an AI capable of generating items by combining images of items the user wants to design with images of natural objects.

[0009] In the invention described in Patent Document 1 above, the learning model (first generator 21) is a machine learning model in which multiple images x of natural objects are used as explanatory variables and multiple images y of artificial objects are used as the objective variable. In the learning stage, the parameters of the first generator 21 are optimized so that the first generator 21 generates images y' that are close enough to real artificial objects to deceive the first classifier 23, which is capable of determining whether the images y' generated by the first generator 21 are images of real artificial objects.

[0010] In the invention described in Patent Document 1 above, when an image of a natural object is input to the first generator 21, a design close to that of a real artificial object is obtained by combining the natural object and the artificial object. Thus, the first generator 21 can be said to be excellent as a generation AI. However, while the first generator 21 prevents the generation of unrealistic designs that are far removed from the item the user wants to design, there is a possibility that only safe and unoriginal designs will be obtained.

[0011] In the invention described in Patent Document 1 above, it is conceivable to train the learning model without using the first discriminator 23 in order to obtain unconventional designs. However, in such a case, the AI ​​may generate unrealistic designs that are far removed from the items that the user wants to design. In that case, the use of AI may not be effective for the user in generating ideas. It is desirable to provide a synthetic data generation system and a synthetic data generation program that can present users with designs that are effective in generating ideas.

[0012] Hereinafter, several exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. The following description is intended to illustrate specific examples of the present disclosure and should not be construed as limiting the disclosure. For example, elements such as numerical values, shapes, materials, parts, the location of each part, and the method of connecting each part are merely examples and should not be construed as limiting the disclosure. Furthermore, in the following exemplary embodiments, components not described in separate sections based on the highest-level concepts of the present disclosure are optional and may be provided as needed. The drawings are schematic and are not intended to be to scale. Throughout this specification and the drawings, components having substantially the same function and substantially the same configuration are denoted by the same reference numerals, and redundant descriptions are omitted. Furthermore, components not directly related to an embodiment of the present disclosure are not shown in the drawings.

[0013] <2. First Embodiment> [Example Configuration] First, a synthetic data generation system 1 according to the first embodiment of this disclosure will be described. Figure 1 shows an example of the functional blocks of the synthetic data generation system 1. The synthetic data generation system 1 includes, for example, a data receiving unit 1A, a generation AI unit 1B, a filter unit 1C, reference data 1D, a storage unit 1E, and a list display unit 1F, as shown in Figure 1. The synthetic data generation system 1 corresponds to one specific example of the "synthetic data generation system" according to one embodiment of this disclosure. The generation AI unit 1B corresponds to one specific example of the "first learning model" according to one embodiment of this disclosure. The filter unit 1C corresponds to one specific example of the "second learning model" and "processing unit" according to one embodiment of this disclosure.

[0014] The data receiving unit 1A has an interface capable of receiving input of natural object data Ia and artificial object data Ib. The data receiving unit 1A receives input of natural object data Ia and artificial object data Ib, which are in a common data format. The data receiving unit 1A is capable of outputting the received natural object data Ia and artificial object data Ib to the generation AI unit 1B. Natural object data Ia and artificial object data Ib are image-related data. Natural object data Ia corresponds to one specific example of the "first data" according to one embodiment of this disclosure. Artificial object data Ib corresponds to one specific example of the "second data" according to one embodiment of this disclosure. Natural object corresponds to one specific example of the "first object" according to one embodiment of this disclosure. Artificial object corresponds to one specific example of the "second object" according to one embodiment of this disclosure.

[0015] Natural object data Ia is, for example, natural object 2D image data Ia1, natural object 3D image data Ia2, or natural object 3D model data. Natural objects are tangible objects that exist in nature and are not artificial or man-made, and include, for example, living things such as dogs and cats. Natural object 2D image data Ia1 is 2D image data that includes natural objects, for example, image data obtained by imaging natural objects with a monocular camera. Natural object 3D image data Ia2 is 3D image data that includes natural objects, for example, image data obtained by imaging natural objects with a 3D camera. Natural object 3D model data is 3D model data that includes natural objects, for example, 3D CAD (Computer-Aided Design) data that includes natural objects.

[0016] Artifact data Ib is, for example, artifact 2D image data Ib1, artifact 3D image data Ib2, or artifact 3D model data. An artifact is an object that is artificially manufactured or constructed, including, for example, a vehicle wheel. Artifact 2D image data Ib1 is two-dimensional image data including an artifact, for example, image data obtained by imaging an artifact with a monocular camera. Artifact 3D image data Ia2 is three-dimensional image data including an artifact, for example, image data obtained by imaging an artifact with a 3D camera. Artifact 3D model data is three-dimensional model data including an artifact, for example, 3D CAD data including an artifact.

[0017] The generation AI unit 1B includes a learning model 11. When natural object data Ia and artificial object data Ib are input to the learning model 11, it is possible to generate and output composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The generation AI unit 1B is also possible to output the composite data Ic generated by the learning model 11 to the filter unit 1C. The learning model 11 corresponds to one specific example of the "first learning model" according to one embodiment of the present disclosure. The composite data Ic corresponds to one specific example of the "composite data" according to one embodiment of the present disclosure. The composite data Ic has a common data format with the natural object data Ia and artificial object data Ib, and is data that includes products obtained by multiplying the natural object and the artificial object together. For example, if the natural object is a dog and the artificial object is a vehicle wheel, the product is a combination of the dog and the wheel together.

[0018] As shown in Figure 2, for example, when natural object 2D image data Ia1 and artificial object 2D image data Ib1 are input to the learning model 11, it is possible to generate and output a product 2D image data Ic1 by multiplying the input natural object 2D image data Ia1 and artificial object 2D image data Ib1 together. The learning model 11 is capable of generating and outputting the product 2D image data Ic1 each time natural object 2D image data Ia1 and artificial object 2D image data Ib1 are input.

[0019] As shown in Figure 3, for example, when natural object 3D image data Ia2 and artificial object 3D image data Ib2 are input to the learning model 11, it is possible to generate and output a product 3D image data Ic2 by multiplying the input natural object 3D image data Ia2 and artificial object 3D image data Ib2 together. The learning model 11 is capable of generating and outputting the product 3D image data Ic2 each time natural object 3D image data Ia2 and artificial object 3D image data Ib2 are input.

[0020] As shown in Figure 4, for example, when natural object 3D model data Ia3 and artificial object 3D model data Ib3 are input to the learning model 11, it is possible to generate and output product 3D model data Ic3 by multiplying the input natural object 3D model data Ia3 and artificial object 3D model data Ib3 together. The learning model 11 is capable of generating and outputting product 3D model data Ic3 each time natural object 3D model data Ia3 and artificial object 3D model data Ib3 are input.

[0021] The learning model 11 is a model that includes, for example, deep learning. The learning model 11 is a model that has been trained using, for example, natural object data Ia_test, artificial object data Ib_test, and synthetic data Ic_test as training data. The natural object data Ia_test is data in the same data format as natural object data Ia and contains natural objects. The artificial object data Ib_test is data in the same data format as artificial object data Ib and contains artificial objects. The synthetic data Ic_test is data in the same data format as synthetic data Ic and contains the products.

[0022] The reference data 1D is a set of data used in the filter unit 1C, and includes, for example, judgment criterion Dref1 and selection criterion Dref2, as shown in Figure 5. Judgment criterion Dref1 corresponds to one specific example of the "judgment criterion" according to one embodiment of the present disclosure. Selection criterion Dref2 corresponds to one specific example of the "selection criterion" according to one embodiment of the present disclosure.

[0023] Criterion Dref1 includes terms that indicate whether the product included in synthetic data Ic is artificial or not, such as "Car Wheel" and "Not Car Wheel". Selection Criterion Dref2 includes a first selection criterion that indicates the product included in synthetic data Ic is not artificial, and a second selection criterion that indicates the product included in synthetic data Ic is artificial. Selection Criterion Dref2 includes, for example, a threshold score for "Not Car Wheel" (e.g., 55% or less) as the first selection criterion, and a threshold score for "Car Wheel" (e.g., 70% or less) as the second selection criterion.

[0024] The filter unit 1C is capable of filtering the synthesized data Ic obtained by the generation AI unit 1B based on the reference data 1D. When the filter unit 1C finds that the synthesized data Ic satisfies the conditions indicated by the reference data 1D as a result of filtering, it is possible to store the synthesized data Ic in association with the filtering result (judgment result Dx1) in the storage unit 1E. The filter unit 1C includes a learning model 12. For example, as shown in Figure 6, when the learning model 12 receives synthesized data Ic and judgment criterion Dref1 as input, it is possible to output a judgment result Dx1 which is a judgment made on the input synthesized data Ic using judgment criterion Dref1. The learning model 12 corresponds to one specific example of the "second learning model" according to one embodiment of the present disclosure. The judgment result Dx1 corresponds to one specific example of the "judgment result" according to one embodiment of the present disclosure. The judgment result Dx1 includes, for example, the score value of "Not Car Wheel" in the synthesized data Ic and the score value of "Car Wheel" in the synthesized data Ic.

[0025] The learning model 12 is a model that includes, for example, CLIP (Contrastive Language-Image Pretraining). The learning model 12 is a model that has been trained using, for example, the synthetic data Ic_test, the judgment criterion Dref1, and the judgment result Dx1_test as training data. The synthetic data Ic_test is data in the same data format as the synthetic data Ic and includes the generated data. The judgment result Dx1_test is data in the same data format as the judgment result Dx1 and includes, for example, the CLIP score value.

[0026] The filter unit 1C is capable of determining whether the judgment result Dx1 obtained from the learning model 12 satisfies the selection criterion Dref2. If the judgment result Dx1 satisfies the selection criterion Dref2, the filter unit 1C can store the synthesized data Ic in association with the judgment result Dx1 in the storage unit 1E. Each time synthesized data Ic is input, the filter unit 1C performs filtering based on the reference data 1D and can determine whether the judgment result Dx1 satisfies the selection criterion Dref2. In this way, the filter unit 1C can store multiple synthesized data Ic that satisfy the selection criterion Dref2 in the storage unit 1E.

[0027] The filter unit 1C is capable of outputting multiple composite data Ic that satisfy the selection criterion Dref2, stored in the storage unit 1E, as a composite data list IcL to the list display unit 1F. For example, the filter unit 1C can generate thumbnail data for each composite data Ic that satisfies the selection criterion Dref2, read from the storage unit 1E, and output the composite data list IcL, which includes the generated thumbnail data, to the list display unit 1F. In this way, the filter unit 1C is capable of performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref2.

[0028] The filter unit 1C is capable of outputting the composite data list IcL to the list display unit 1F at predetermined timings. For example, the filter unit 1C can output the composite data list IcL to the list display unit 1F when requested by a user of the composite generation system 1, or whenever composite data Ic that satisfies the selection criterion Dref2 is stored in the storage unit 1E.

[0029] When a composite data list IcL is input to the list display unit 1F, it is possible to generate an interface IF1 that includes an image listing the multiple composite data Ic contained in the input composite data list IcL, as shown in Figure 7. Interface IF1 has a function that allows for enlargement of the composite data Ic selected by the user. Interface IF1 also has a function that allows for the generation of a selection flag for the composite data Ic selected by the user, and to store the selection flag in the storage unit 1E, associating it with the composite data Ic selected by the user from among the multiple composite data Ic contained in the storage unit 1E.

[0030] In the synthetic data generation system 1, the filter unit 1C may be capable of generating an interface IF2 that can adjust the score thresholds for the first selection criterion ("Car Wheel") and the score thresholds for the second selection criterion ("Not Car Wheel"), for example, as shown in Figure 8. In this case, the interface IF2 has a function that can adjust the score thresholds for the first selection criterion ("Car Wheel") and the score thresholds for the second selection criterion ("Not Car Wheel") in response to user input. When the score thresholds for the first selection criterion ("Car Wheel") and the score thresholds for the second selection criterion ("Not Car Wheel") are adjusted, the filter unit 1C can update the reference data 1D with the adjusted thresholds.

[0031] [effect] Next, we will explain the effects of the synthetic data generation system 1.

[0032] In this embodiment, when natural object data Ia and artificial object data Ib1 are input to the learning model 11, composite data Ic is output by multiplying the input natural object data Ia and artificial object data Ib1 together. When the composite data Ic and judgment criterion Dref1 are input to the learning model 12, a judgment result Dx1 is output, which is the result of the input composite data Ic being judged by judgment criterion Dref1. Then, in the filter unit 1C, it is determined whether the judgment result Dx1 satisfies the selection criterion Dref2, and data processing is performed to display a list of multiple composite data Ics that satisfy the selection criterion Dref2. This makes it possible to eliminate unrealistic designs that are far removed from the artificial object (item) that the user wants to design, as well as safe and unoriginal designs, before they are displayed in the list. As a result, designs that are effective for idea generation can be presented to the user.

[0033] In this embodiment, the synthesized data Ic is stored in the storage unit 1E in association with the judgment result Dx1. This allows the newly obtained synthesized data Ic to be displayed in a list together with multiple synthesized data Ic from the past in which the judgment result Dx1 met the selection criterion Dref2. As a result, multiple designs (synthetic data Ic) that are effective for idea generation can be presented to the user.

[0034] <3. Modified Examples of the First Embodiment> Next, a modified example of the synthetic data generation system 1 will be described.

[0035] [Differentiation A] Figure 9 shows a modified example of the functional block of the synthetic data generation system 1. The synthetic data generation system 1 may be capable of receiving multiple (N) natural object data Ia at the data reception unit 1A, as shown in Figure 9. Multiple (N) natural object data Ia correspond to one specific example of the "multiple first data" according to one embodiment of the present disclosure. Each of the multiple (N) natural object data Ia contains a different natural object. When two natural object data Ia are input to the data reception unit 1A, in the two natural object data Ia input to the data reception unit 1A, the natural object included in one natural object data Ia is, for example, a dog, and the natural object included in the other natural object data Ia is, for example, a cat.

[0036] In this modified example, the data reception unit 1A has an interface capable of receiving input of multiple (N) natural object data Ia and one artificial object data Ib. The data reception unit 1A can input the received multiple (N) natural object data Ia and one artificial object data Ib into the generation AI unit 1B. In the generation AI unit 1B, when the learning model 11 receives multiple (N) natural object data Ia and one artificial object data Ib as input, it is possible to generate and output composite data Ic by multiplying the multiple (N) natural object data Ia and the one artificial object data Ib together. In this modified example, the learning model 11 is a model that has been trained using, for example, multiple (N) natural object data Ia_test, one artificial object data Ib_test, and one composite data Ic_test as training data.

[0037] As shown in Figure 10, for example, when the learning model 11 receives multiple (N) natural object 2D image data Ia1 and one artificial object 2D image data Ib1 as input, it is possible to generate and output a product 2D image data Ic1 by multiplying the input multiple (N) natural object 2D image data Ia1 and one artificial object 2D image data Ib1 together. The learning model 11 is capable of generating and outputting the product 2D image data Ic1 each time multiple (N) natural object 2D image data Ia1 and artificial object 2D image data Ib1 are input.

[0038] As shown in Figure 11, for example, when the learning model 11 receives multiple (N) natural object 3D image data Ia2 and one artificial object 3D image data Ib2 as input, it is possible to generate and output a product 3D image data Ic2 by multiplying the input multiple (N) natural object 3D image data Ia2 and one artificial object 3D image data Ib2 together. The learning model 11 is capable of generating and outputting the product 3D image data Ic2 each time multiple (N) natural object 3D image data Ia2 and one artificial object 3D image data Ib2 are input.

[0039] As shown in Figure 12, for example, when the learning model 11 receives multiple (N) natural object 3D model data Ia3 and one artificial object 3D model data Ib3 as input, it is possible to generate and output a product 3D model data Ic3 by multiplying the input multiple (N) natural object 3D model data Ia3 and one artificial object 3D model data Ib3 together. The learning model 11 is capable of generating and outputting the product 3D model data Ic3 each time multiple (N) natural object 3D model data Ia3 and one artificial object 3D model data Ib3 are input.

[0040] The product included in the synthesized data Ic is a product of multiple natural objects included in multiple (N) natural object data Ia and artificial objects included in the artificial object data Ib. When two natural object data Ia are input to the generation AI unit 1B, suppose one of the two natural object data Ia input to the generation AI unit 1B includes, for example, a dog, and the other natural object data Ia includes, for example, a cat. Furthermore, suppose the artificial object data Ib includes a vehicle wheel. In this case, the product included in the synthesized data Ic is a product of the dog, cat, and wheel.

[0041] In this modified example, the generation AI unit 1B receives multiple (N) natural object data Ia and one artificial object data Ib as input. This allows for the effective creation of more unconventional designs while presenting the user with designs that are effective for generating ideas.

[0042] [Variation B] Figure 13 shows a modified example of the functional block of the synthetic data generation system 1. The synthetic data generation system 1 may have an interface in the data reception unit 1A that can receive natural object data Ta and artificial object data Tb, for example, as shown in Figure 13. The data reception unit 1A receives natural object data Ta and artificial object data Tb in a common data format. The data reception unit 1A is capable of inputting the received natural object data Ta and artificial object data Tb into the data conversion unit 1G. The natural object data Ta corresponds to a specific example of the "first character data" according to one embodiment of the present disclosure. The artificial object data Tb corresponds to a specific example of the "second character data" according to one embodiment of the present disclosure.

[0043] Natural object data Ta is character data representing a natural object. Examples of character data representing natural objects include DOG or CAT. Artifact data Tb is character data representing an artifact. Examples of character data representing an artifact include WHEEL.

[0044] In this modified example, the synthetic data generation system 1 further comprises a data conversion unit 1G and conversion data 1H, as shown in Figure 13, for example. When natural object data Ta and artificial object data Tb are input to the data conversion unit 1G, it is possible to convert the input natural object data Ta and artificial object data Tb into natural object data Ia and artificial object data Ib using the conversion data 1H. The data conversion unit 1G is also capable of outputting the natural object data Ia and artificial object data Ib obtained by the conversion to the generation AI unit 1B.

[0045] The conversion data 1H contains, for example, natural object data Ia associated with natural object data Ta, and artificial object data Ib associated with artificial object data Tb. When natural object data Ta is input to the data conversion unit 1G, it is possible to extract natural object data Ia corresponding to the input natural object data Ta from the conversion data 1H. When artificial object data Tb is input to the data conversion unit 1G, it is possible to extract artificial object data Ib corresponding to the input artificial object data Tb from the conversion data 1H.

[0046] As shown in Figure 14, for example, when natural object character data Ta1 and artificial object character data Tb1 are input to the data conversion unit 1G, it is possible to convert the input natural object character data Ta1 and artificial object character data Tb1 into natural object 2D image data Ia1 and artificial object 2D image data Ib1 using conversion data 1H. Natural object character data Ta1 is character data that represents a natural object. Artificial object character data Tb1 is character data that represents an artificial object. Each time natural object character data Ta1 and artificial object character data Tb1 are input to the data conversion unit 1G, it is possible to convert the input natural object character data Ta1 and artificial object character data Tb1 into natural object 2D image data Ia1 and artificial object 2D image data Ib1 using conversion data 1H.

[0047] The data conversion unit 1G may, for example, as shown in Figure 15, be capable of converting the input natural object character data Ta1 and artificial object character data Tb1 into natural object 3D image data Ia2 and artificial object 3D image data Ib2 using conversion data 1H when natural object character data Ta1 and artificial object character data Tb1 are input. The data conversion unit 1G may also be capable of converting the input natural object character data Ta1 and artificial object character data Tb1 into natural object 3D image data Ia2 and artificial object 3D image data Ib2 using conversion data 1H each time natural object character data Ta1 and artificial object character data Tb1 are input.

[0048] The data conversion unit 1G may, for example, as shown in Figure 16, be capable of converting the input natural object character data Ta1 and artificial object character data Tb1 into natural object 3D model data Ia3 and artificial object 3D model data Ib3 using conversion data 1H when natural object character data Ta1 and artificial object character data Tb1 are input. The data conversion unit 1G may also be capable of converting the input natural object character data Ta1 and artificial object character data Tb1 into natural object 3D model data Ia3 and artificial object 3D model data Ib3 using conversion data 1H each time natural object character data Ta1 and artificial object character data Tb1 are input.

[0049] In this modified example, the natural object data Ta and artificial object data Tb received by the data reception unit 1A are converted into natural object data Ia and artificial object data Ib by the data conversion unit 1G. The converted natural object data Ia and artificial object data Ib are then input to the generation AI unit 1B. As a result, the user can obtain synthesized data Ic simply by inputting text data. Consequently, the effort required for data input by the user is reduced, while the user can be presented with designs that are effective for idea generation.

[0050] [Differentiation C] Figure 17 shows one modified example of the functional block of the synthetic data generation system 1. In modified example B, the synthetic data generation system 1 may be capable of converting natural object data Ta into multiple (N) natural object data Ia in the data conversion unit 1G using conversion data 1H, as shown in Figure 17. In this case, the data conversion unit 1G is capable of outputting the multiple (N) natural object data Ia and one artificial object data Ib obtained by the conversion to the generation AI unit 1B.

[0051] Multiple (N) natural object data Ia each contain different natural objects. When two natural object data Ia are input to the data conversion unit 1G, one of the two natural object data Ia input to the data conversion unit 1G may contain, for example, a dog, and the other natural object data Ia may contain, for example, a cat.

[0052] The data conversion unit 1G can, for example, when natural object character data Ta1 and artificial object character data Tb1 are input, use conversion data 1H to convert the input natural object character data Ta1 and artificial object character data Tb1 into multiple (N) natural object 2D image data Ia1 and one artificial object 2D image data Ib1. The data conversion unit 1G can convert the input natural object character data Ta1 and artificial object character data Tb1 into multiple (N) natural object 2D image data Ia1 and one artificial object 2D image data Ib1 using conversion data 1H each time natural object character data Ta1 and artificial object character data Tb1 are input.

[0053] The data conversion unit 1G may, for example, when natural object character data Ta1 and artificial object character data Tb1 are input, use conversion data 1H to convert the input natural object character data Ta1 and artificial object character data Tb1 into multiple (N) natural object 3D image data Ia2 and one artificial object 3D image data Ib2. The data conversion unit 1G may also, each time natural object character data Ta1 and artificial object character data Tb1 are input, use conversion data 1H to convert the input natural object character data Ta1 and artificial object character data Tb1 into multiple (N) natural object 3D image data Ia2 and one artificial object 3D image data Ib2.

[0054] The data conversion unit 1G may, for example, be capable of converting the input natural object character data Ta1 and artificial object character data Tb1 into multiple (N) natural object 3D model data Ia3 and one artificial object 3D model data Ib3 using conversion data 1H when natural object character data Ta1 and artificial object character data Tb1 are input. The data conversion unit 1G may also be capable of converting the input natural object character data Ta1 and artificial object character data Tb1 into multiple (N) natural object 3D model data Ia3 and one artificial object 3D model data Ib3 using conversion data 1H each time natural object character data Ta1 and artificial object character data Tb1 are input.

[0055] In this modified example, the generating AI unit 1B (learning model 11) is capable of generating and outputting composite data Ic, which is obtained by multiplying the multiple (N) natural object data Ia and the one artificial object data Ib together, when multiple (N) natural object data Ia and one artificial object data Ib are input. In this modified example, the learning model 11 is a model that has been trained using, for example, multiple (N) natural object data Ia_test, one artificial object data Ib_test, and one composite data Ic_test as training data.

[0056] In this modified example, the data conversion unit 1G uses the conversion data 1H to convert the natural object data Ta into multiple (N) natural object data Ia. This makes it possible to effectively create more unconventional designs while presenting users with designs that are effective in generating ideas.

[0057] [Differentiation D] Figure 18 shows one modified example of the functional block of the synthetic data generation system 1. In modified example C, the synthetic data generation system 1 may be capable of receiving multiple (N) natural object data Ta at the data reception unit 1A, as shown in Figure 18. In this case, the data reception unit 1A is capable of outputting the received multiple (N) natural object data Ta and one artificial object data Tb to the data conversion unit 1G. In this modified example, when the data conversion unit 1G receives multiple (N) natural object data Ta and one artificial object data Tb as input, it is possible to convert the multiple (N) natural object data Ta and one artificial object data Tb into multiple (N) natural object data Ia and one artificial object data Ib using conversion data 1H.

[0058] Thus, although this modified version differs from modified version C in that multiple (N) natural object data Ta are input to the data reception unit 1A, it is possible to effectively create more unconventional designs while presenting users with designs that are effective in generating ideas, similar to modified version C.

[0059] <4. Application Examples> Next, examples of applications of the synthetic data generation system 1 according to the first embodiment and modified versions A to D will be described.

[0060] Figure 19 shows an example of the internal configuration of an information processing device when the composite data generation system 1 according to the first embodiment is applied to a single information processing device. The composite data generation system 1 according to this application example includes, for example, an input unit 10, a storage unit 20, a storage unit 30, a control unit 40, and a display unit 50, as shown in Figure 19.

[0061] The input unit 10 is capable of performing the same functions as the data reception unit 1A. The input unit 10 has an interface capable of receiving input of natural object data Ia and artificial object data Ib. The input unit 10 receives input of natural object data Ia and artificial object data Ib, which are in a common data format. The input unit 10 is capable of outputting the received natural object data Ia and artificial object data Ib to the control unit 40.

[0062] The storage unit 20 is a non-temporary tangible recording medium. The storage unit 20 is composed of, for example, non-volatile memory, such as EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, or resistive random-access memory. The storage unit 20 stores, for example, program 21 and learning models 11 and 12, as shown in Figure 19. Program 21 is a program that causes the control unit 40 to execute a series of procedures for realizing list display in the synthetic data generation system 1 according to this application example. Program 21 corresponds to one specific example of the "synthetic data generation program" according to one embodiment of this disclosure.

[0063] The storage unit 30 is composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory. The storage unit 30 stores reference data 31, for example, as shown in Figure 19. The reference data 31 includes judgment criterion Dref1 and selection criterion Dref2. The storage unit 30 stores composite data 32 and judgment result 33. The composite data 32 includes composite data Ic generated by the control unit 40. The generated composite data Ic is added to the composite data 32 each time the control unit 40 generates composite data Ic. The judgment result 33 includes judgment result Dx1. The generated judgment result Dx1 is added to the judgment result 33 each time the control unit 40 generates judgment result Dx1. In the composite data 32 and judgment result 33, the composite data Ic and judgment result Dx1 are associated with each other.

[0064] The control unit 40 can execute a series of procedures to realize list display in the synthetic data generation system 1 according to this application example once the program 21 is loaded into the control unit 40. The control unit 40 can implement functions similar to those of the generation AI unit 1B using the learning model 11. The control unit 40 can implement functions similar to those of the filter unit 1C using the learning model 12 and reference data 31.

[0065] When natural object data Ia and artificial object data Ib are input to the control unit 40, the control unit 40 can input the input natural object data Ia and artificial object data Ib to the learning model 11. In response to the input of natural object data Ia and artificial object data Ib to the learning model 11, the control unit 40 can acquire synthesized data Ic from the learning model 11.

[0066] The control unit 40 is capable of filtering the synthesized data Ic acquired from the learning model 11 based on the reference data 1D. When the combined data Ic satisfies the conditions indicated by the reference data 1D as a result of filtering, the control unit 40 can store the combined data Ic in the storage unit 30 (composite data 32, judgment result 33) in association with the filtering result (judgment result Dx1). The control unit 40 can input the combined data Ic and judgment criterion Dref1 to the learning model 12. In response to the input of the combined data Ic and judgment criterion Dref1 to the learning model 12, the control unit 40 can acquire the judgment result Dx1 from the learning model 12.

[0067] The control unit 40 is capable of determining whether the judgment result Dx1 obtained from the learning model 12 satisfies the selection criterion Dref2. If the judgment result Dx1 satisfies the selection criterion Dref2, the control unit 40 can store the synthesized data Ic in the storage unit 30 (synthetic data 32, judgment result 33) in association with the judgment result Dx1. Each time synthesized data Ic is obtained, the control unit 40 can perform filtering based on the reference data 1D and determine whether the judgment result Dx1 satisfies the selection criterion Dref2. In this way, the control unit 40 can store multiple synthesized data Ic that satisfy the selection criterion Dref2 in the storage unit 30.

[0068] The control unit 40 is capable of outputting a list of composite data Ic stored in the storage unit 30 that satisfy the selection criterion Dref2 to the display unit 50 as a composite data list IcL. For example, the control unit 40 can generate thumbnail data for each composite data Ic that satisfies the selection criterion Dref2 read from the storage unit 30, and output the composite data list IcL containing the generated thumbnail data to the display unit 50. In this way, the control unit 40 is capable of performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref2.

[0069] The control unit 40 is capable of outputting the composite data list IcL to the display unit 50 at predetermined timings. For example, the control unit 40 can output the composite data list IcL to the display unit 50 when requested by a user of the composite generation system 1, or whenever composite data Ic that satisfies the selection criterion Dref2 is stored in the storage unit 30.

[0070] The display unit 50 is capable of displaying a list of multiple composite data Ic (composite data list IcL) generated by the control unit 40. When the composite data list IcL is input to the display unit 50, it is capable of generating an interface IF1. The display unit 50 is configured to include, for example, a liquid crystal panel or an organic EL panel. In the interface IF1, the display unit 50 is capable of displaying, for example, a composite data Ic selected by the user in an enlarged view. In the interface IF1, the display unit 50 is capable of generating a selection flag for, for example, the composite data Ic selected by the user and outputting the generated selection flag to the control unit 40. At this time, the control unit 40 is capable of storing the selection flag in the storage unit 30, associating it with the composite data Ic selected by the user from among the multiple composite data Ic contained in the storage unit 1E.

[0071] The display unit 50 may be capable of generating interface IF2. The display unit 50 can output the adjusted threshold values ​​to the control unit 40 when the threshold values ​​for the first selection criterion ("Car Wheel") or the second selection criterion ("Not Car Wheel") are adjusted. The control unit 40 can store the adjusted threshold values ​​in the storage unit 30 (reference data 31).

[0072] Next, the list display procedure in the synthetic data generation system 1 related to this application example will be described. Figure 20 shows an example of the list display procedure in the synthetic data generation system 1 related to this application example.

[0073] The control unit 40 first receives natural object data Ia and artificial object data Ib as input (step S101). The control unit 40 inputs the received natural object data Ia and artificial object data Ib into the learning model 11. The control unit 40 uses the learning model 11 to obtain composite data Ic (step S102). The control unit 40 inputs the composite data Ic and the judgment criterion Dref1 into the learning model 12. The control unit 40 uses the learning model 12 to obtain the judgment result Dx1 (step S103).

[0074] The control unit 40 determines whether the determination result Dx1 satisfies the selection criterion Dref2 (step S104). If the determination result Dx1 satisfies the selection criterion Dref2, the control unit 40 stores the synthesized data Ic in association with the determination result Dx1 in the storage unit 30 (step S105). Each time synthesized data Ic is obtained, the control unit 40 uses the learning model 12 to acquire the determination result Dx1, and if the determination result Dx1 satisfies the selection criterion Dref2, it stores the synthesized data Ic in association with the determination result Dx1 in the storage unit 30. In this way, the control unit 40 stores multiple synthesized data Ic that satisfy the selection criterion Dref2 in the storage unit 30.

[0075] The control unit 40 outputs a list of composite data Ic that satisfy the selection criterion Dref2, stored in the storage unit 30, to the display unit 50 as a composite data list IcL. The display unit 50 generates an interface IF1 that includes an image listing the multiple composite data Ic included in the composite data list IcL. As a result, the display unit 50 displays a list of the multiple composite data Ic that satisfy the selection criterion Dref2 (step S106). In this way, the list display in the composite data generation system 1 according to this application example is performed.

[0076] In this application example, the synthetic data generation system 1 according to the first embodiment is implemented by a single information processing device. As a result, the user can input natural object data Ia and artificial object data Ib into a single information processing device and obtain only the design (synthetic data Ic) that is effective for idea generation.

[0077] In this application example, a single information processing device may be capable of receiving input from multiple (N) natural object data Ia and one artificial object data Ib. In this case, by inputting multiple (N) natural object data Ia and one artificial object data Ib into a single information processing device, the user can obtain only the more unconventional designs (synthetic data Ic) that are effective for idea generation.

[0078] In this application example, for example, as shown in Figure 21, a single information processing device may be configured to receive input of natural object data Ta and artificial object data Tb. In this case, the storage unit 30 further stores natural object data 34 and artificial object data 35. Natural object data 34 stores natural object data Ia in association with natural object data Ta. Artificial object data 35 stores artificial object data Ib in association with artificial object data Tb.

[0079] The input unit 10 has an interface capable of receiving natural object data Ta and artificial object data Tb. The input unit 10 receives natural object data Ta and artificial object data Tb in a common data format. The input unit 10 can input the received natural object data Ta and artificial object data Tb to the control unit 40.

[0080] When natural object data Ta and artificial object data Tb are input to the control unit 40, it is possible to extract natural object data Ia corresponding to the natural object data Ta and artificial object data Ib corresponding to the artificial object data Tb from the storage unit 30 (natural object data 34, artificial object data 35). The control unit 40 is also possible to input the natural object data Ia and artificial object data Ib extracted from the storage unit 30 into the learning model 11.

[0081] Next, a modified example of the list display procedure in the synthetic data generation system 1 according to this application example will be described. Figure 22 shows a modified example of the list display procedure in the synthetic data generation system 1 according to this application example.

[0082] The control unit 40 first receives input of natural object data Ta and artificial object data Tb (step S201). The control unit 40 uses the storage unit 30 (natural object data 34, artificial object data 35) to convert the received natural object data Ta and artificial object data Tb into natural object data Ia and artificial object data Ib (step S202). The control unit 40 inputs the natural object data Ia and artificial object data Ib obtained by the conversion into the learning model 11. The control unit 40 uses the learning model 11 to obtain composite data Ic (step S203). The control unit 40 inputs the composite data Ic and the judgment criterion Dref1 into the learning model 12. The control unit 40 uses the learning model 12 to obtain the judgment result Dx1 (step S204).

[0083] The control unit 40 determines whether the determination result Dx1 satisfies the selection criterion Dref2 (step S205). If the determination result Dx1 satisfies the selection criterion Dref2, the control unit 40 stores the synthesized data Ic in association with the determination result Dx1 in the storage unit 30 (step S206). Each time synthesized data Ic is obtained, the control unit 40 uses the learning model 12 to acquire the determination result Dx1, and if the determination result Dx1 satisfies the selection criterion Dref2, it stores the synthesized data Ic in association with the determination result Dx1 in the storage unit 30. In this way, the control unit 40 stores multiple synthesized data Ic that satisfy the selection criterion Dref2 in the storage unit 30.

[0084] The control unit 40 outputs multiple composite data Ic that satisfy the selection criterion Dref2, stored in the storage unit 30, to the display unit 50 as a composite data list IcL. The display unit 50 generates an interface IF1 that includes an image listing the multiple composite data Ic included in the composite data list IcL. As a result, the display unit 50 displays a list of multiple composite data Ic that satisfy the selection criterion Dref2 (step S207). In this way, the list display in the composite data generation system 1 according to this application example is performed.

[0085] In this application example, the synthetic data generation system 1 according to the first embodiment is implemented by a single information processing device. As a result, the user can input natural object data Ta and artificial object data Tb into a single information processing device and obtain only the design (synthetic data Ic) that is effective for idea generation.

[0086] In this application example, for example, as shown in Figure 23, the composite data generation system 1 shown in Figure 19 may be implemented using terminal devices 100 and server devices 200 connected via a communication network 300. The communication network 300 may be, for example, a wired LAN (Local Area Network) such as Ethernet, a wireless LAN such as Wi-Fi, or a mobile phone line.

[0087] The terminal device 100 includes, for example, an input unit 10, a display unit 50, a storage unit 110, a control unit 120, and a communication unit 130. The communication unit 130 includes an interface capable of communicating with the server device 200 via a communication network 300. The storage unit 110 is a non-temporary tangible recording medium. The storage unit 110 is composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory. The storage unit 110 stores a program 111. The program 111 includes a series of procedures for transmitting natural object data Ia and artificial object data Ib received by the input unit 10 to the server device 200, and for performing data processing to list and display multiple composite data Ic that satisfy the selection criterion Dref2, which are included in the composite data list IcL received from the server device 200.

[0088] Once program 111 is loaded into the control unit 120, the control unit 120 is able to transmit the natural object data Ia and artificial object data Ib received by the input unit 10 to the server device 200. Once program 111 is loaded into the control unit 120, the control unit 120 is able to perform a series of procedures for data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref2, which are included in the composite data list IcL received from the server device 200, on the display unit 50. The display unit 50 is able to display a list of multiple composite data Ic included in the composite data list IcL.

[0089] The server device 200 includes, for example, a communication unit 210, storage units 220, 230, and a control unit 240. The communication unit 210 includes an interface capable of communicating with the terminal device 100 via a communication network 300. The storage units 220, 230 are non-temporary tangible recording media. The storage units 220, 230 are composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory.

[0090] The storage unit 220 stores the program 221 and the learning models 11 and 12. The program 221 includes a series of procedures for receiving natural object data Ia and artificial object data Ib from the terminal device 100, and for generating a composite data list IcL using the natural object data Ia and artificial object data Ib received from the terminal device 100. The storage unit 230 stores the reference data 31. The storage unit 230 also stores the composite data 32 and the judgment result 33.

[0091] Once the program 221 is loaded into the control unit 240, it becomes possible to perform a series of procedures for receiving natural object data Ia and artificial object data Ib from the terminal device 100, and for generating a composite data list IcL using the natural object data Ia and artificial object data Ib received from the terminal device 100.

[0092] Thus, when the composite data generation system 1 shown in Figure 19 is implemented using terminal devices 100 and server devices 200 connected via a communication network 300, relatively low-spec terminal devices can be used as terminal devices 100. Furthermore, since multiple terminal devices 100 can share the server device 200, the number of terminal devices 100 can be increased at low cost.

[0093] In this application example, for example, as shown in Figure 24, the composite data generation system 1 shown in Figure 21 may be implemented using terminal devices 100 and server devices 200 connected via a communication network 300. In the server device 200, the storage unit 230 stores reference data 31, natural object data 34, and artificial object data 35. The storage unit 230 also stores composite data 32 and judgment results 33.

[0094] Thus, when the composite data generation system 1 shown in Figure 21 is implemented using terminal devices 100 and server devices 200 connected via a communication network 300, relatively low-spec terminal devices can be used as terminal devices 100. Furthermore, since multiple terminal devices 100 can share the server device 200, the number of terminal devices 100 can be increased at low cost.

[0095] <5. Second Embodiment> [Example Configuration] Next, a synthetic data generation system 2 according to a second embodiment of the present disclosure will be described. Figure 25 shows an example of the functional blocks of the synthetic data generation system 2. The synthetic data generation system 2 includes, for example, a data receiving unit 2A, a generation AI unit 2B, a selection flag registration unit 2C, a storage unit 2D, a data receiving unit 2E, a generation AI unit 2F, a filter unit 2G, reference data 2H, and a list display unit 2I, as shown in Figure 25.

[0096] The synthetic data generation system 2 corresponds to a specific example of the "synthetic data generation system" according to one embodiment of the present disclosure. The generation AI unit 2B corresponds to a specific example of the "first learning model" according to one embodiment of the present disclosure. The selection flag registration unit 2C corresponds to a specific example of the "processing unit" according to one embodiment of the present disclosure. The filter unit 2G corresponds to a specific example of the "second learning model" and "processing unit" according to one embodiment of the present disclosure.

[0097] The data reception unit 2A has an interface capable of receiving input of a natural object data set Ia_set and artificial object data Ib. The natural object data set Ia_set includes multiple natural object data Ia. Each of the multiple natural object data Ia included in the natural object data set Ia_set contains a natural object that is different from each other in at least one of the following aspects: individual, angle, and scale. The multiple natural object data Ia included in the natural object data set Ia_set may also contain a natural object that is exactly the same in all of the natural object data Ia in terms of individual, angle, and scale. The data reception unit 2A receives input of natural object data Ia and artificial object data Ib in a common data format. The data reception unit 2A is capable of inputting the received natural object data set Ia_set and artificial object data Ib to the generation AI unit 2B.

[0098] When the generation AI unit 2B receives the natural object data set Ia_set and the artificial object data Ib as input, it inputs multiple natural object data Ia from the input natural object data set Ia_set one by one into the learning model 11, and can also input the artificial object data Ib into the learning model 11 at the same time as inputting the natural object data Ia. When the learning model 11 receives the natural object data Ia and the artificial object data Ib as input, it can generate and output the composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The learning model 11 can generate and output the composite data Ic each time natural object data Ia and artificial object data Ib are input.

[0099] The selection flag registration unit 2C can generate an interface IF3 that includes an image listing multiple composite data Ic obtained by the generation AI unit 2B, as shown in Figure 26. Interface IF3 can enlarge and display a composite data Ic specified by the user, for example. Interface IF3 can display a composite data Ic selected by the user (composite data Ic_select) in a way that distinguishes it from composite data Ic not selected by the user (composite data Ic_Non-select), as shown in Figure 27.

[0100] Interface IF3 is capable of generating a selection flag for the synthesized data Ic_select and a non-selection flag for the synthesized data Ic_Non-select. Interface IF3 can store the flag data set Flg_set, which includes one or more selection flags and one or more non-selection flags, in the storage unit 2D together with the multiple synthesized data Ic obtained by the generation AI unit 2B.

[0101] Interface IF3 is capable of storing one or more synthetic data Ic_Non-select as criterion Dref3 in reference data 2H. Interface IF3 may also be capable of storing the features of one or more synthetic data Ic_Non-select as criterion Dref3 in reference data 2H. Criterion Dref3 corresponds to one specific example of the "criterion" according to one embodiment of this disclosure.

[0102] The reference data 2H is a data set used in the filter unit 2G, and includes, for example, the judgment criterion Dref3 and the selection criterion Dref4, as shown in Figure 28. The judgment criterion Dref3 corresponds to one specific example of the "judgment criterion" according to one embodiment of the present disclosure. The selection criterion Dref4 corresponds to one specific example of the "selection criterion" according to one embodiment of the present disclosure. The judgment criterion Dref3 includes one or more composite data Ic_Non-select or the feature quantities of one or more composite data Ic_Non-select. The selection criterion Dref4 includes a threshold for the similarity (e.g., 55% or more and 70% or less) generated by the filter unit 2G, as described below.

[0103] The data reception unit 2E has an interface capable of receiving input of natural object data Ia and artificial object data Ib. The data reception unit 2E receives input of natural object data Ia and artificial object data Ib, which are in a common data format. The data reception unit 2E is capable of outputting the received natural object data Ia and artificial object data Ib to the generation AI unit 2F. The generation AI unit 2F includes a learning model 11. When the learning model 11 receives input of natural object data Ia and artificial object data Ib, it is capable of generating and outputting composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The generation AI unit 2F is capable of outputting the composite data Ic generated by the learning model 11 to the filter unit 2G.

[0104] The filter unit 2G includes a learning model 13. When the learning model 13 receives synthetic data Ic generated by the generation AI unit 2F and one or more synthetic data Ic_Non-select (judgment criterion Dref3) as input, it is capable of generating and outputting a judgment result Dx2 which is a judgment made on the input synthetic data Ic using the input judgment criterion Dref3. The learning model 13 corresponds to one specific example of the "second learning model" according to one embodiment of this disclosure. The judgment result Dx2 is data indicating the similarity (first similarity) between the synthetic data Ic generated by the generation AI unit 2F and one or more synthetic data Ic_Non-select (judgment criterion Dref3). The judgment result Dx2 may also be data indicating the similarity (second similarity) between the feature quantities of the synthetic data Ic generated by the generation AI unit 2F and one or more feature quantities (judgment criterion Dref3) of the synthetic data Ic_Non-select.

[0105] The filter unit 2G is capable of determining whether the judgment result Dx2 obtained from the learning model 13 satisfies the selection criterion Dref4. If the judgment result Dx2 satisfies the selection criterion Dref4, the filter unit 2G can store the synthesized data Ic in association with the judgment result Dx2 in the storage unit 2D. Each time synthesized data Ic is input, the filter unit 2G performs filtering based on the reference data 2H and can determine whether the judgment result Dx2 satisfies the selection criterion Dref4. In this way, the filter unit 2G is capable of storing multiple synthesized data Ic that satisfy the selection criterion Dref4 in the storage unit 2D.

[0106] The filter unit 2G is capable of outputting multiple composite data Ic that satisfy the selection criterion Dref4, stored in the storage unit 2D, as a composite data list IcL to the list display unit 2I. For example, the filter unit 2G can generate thumbnail data for each composite data Ic that satisfies the selection criterion Dref4, read from the storage unit 2D, and output the composite data list IcL, which includes the generated thumbnail data, to the list display unit 2I. In this way, the filter unit 2G is capable of performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref4.

[0107] The filter unit 2G is capable of outputting the composite data list IcL to the list display unit 2I at predetermined timings. For example, the filter unit 2G can output the composite data list IcL to the list display unit 2I when requested by a user of the composite generation system 2, or whenever composite data Ic that satisfies the selection criterion Dref4 is stored in the storage unit 2D.

[0108] When a composite data list IcL is input to the list display unit 2I, it is possible to generate an interface IF4 that includes an image listing the multiple composite data Ic contained in the input composite data list IcL, as shown in Figure 30, for example. Interface IF4 has a function that allows for enlargement of a composite data Ic specified by the user.

[0109] In the synthetic data generation system 2, the filter unit 2G may be capable of generating an interface IF5 that can adjust the threshold values ​​for the first or second similarity, as shown in Figure 31, for example. In this case, the interface IF5 has a function that can adjust the upper limit of the threshold value for the first or second similarity, or the lower limit of the threshold value for the first or second similarity, in response to user input, for example. When the upper limit of the threshold value for the first or second similarity, or the lower limit of the threshold value for the first or second similarity, is adjusted, the filter unit 2G can update the reference data 2H with the adjusted threshold values.

[0110] [effect] Next, we will explain the effects of the synthetic data generation system 2.

[0111] In this embodiment, when natural object data Ia and artificial object data Ib1 are input to the learning model 11, composite data Ic is output by multiplying the input natural object data Ia and artificial object data Ib1 together. When the composite data Ic and judgment criterion Dref3 are input to the learning model 13, judgment result Dx2 is output, which is the result of the input composite data Ic being judged by judgment criterion Dref3. Then, in the filter unit 2G, it is determined whether the judgment result Dx2 satisfies the selection criterion Dref4, and data processing is performed to display a list of multiple composite data Ics that satisfy the selection criterion Dref4. This makes it possible to eliminate unrealistic designs that are far removed from the artificial object (item) that the user wants to design, as well as safe designs that lack originality, before they are displayed in the list. As a result, designs that are effective for idea generation can be presented to the user.

[0112] In this embodiment, the synthesized data Ic is stored in the storage unit 2D in association with the judgment result Dx2. This allows the newly obtained synthesized data Ic to be displayed in a list together with multiple synthesized data Ic from the past in which the judgment result Dx2 met the selection criterion Dref4. As a result, multiple designs (synthetic data Ic) that are effective for idea generation can be presented to the user.

[0113] In this embodiment, one or more composite data Ic_Non-select or their feature quantities are stored in the reference data 2H as the judgment criterion Dref3. This allows for the exclusion of unrealistic designs that are far removed from the artifact (item) the user wants to design, as well as bland, unoriginal designs, before they are displayed in the list. As a result, the user can be presented with designs that are effective in generating ideas.

[0114] In this embodiment, the judgment result Dx2 is data indicating the first similarity or the second similarity described above, and the selection criterion Dref4 includes a threshold for the first similarity or the second similarity described above. This makes it possible to preemptively eliminate unrealistic designs that are far removed from the artifact (item) that the user wants to design, as well as bland, unoriginal designs, before they are displayed in the list. As a result, it is possible to present the user with designs that are effective in generating ideas.

[0115] <6. Modified Version of the Second Embodiment> Next, a modified example of the synthetic data generation system 2 will be described.

[0116] [Differentiation Example E] Figure 32 shows a modified example of the functional block of the synthetic data generation system 2. The synthetic data generation system 2 may be capable of receiving multiple (N) natural object data Ia at the data reception unit 2E, for example, as shown in Figure 32. Each of the multiple (N) natural object data Ia contains a different natural object. When two natural object data Ia are input to the data reception unit 2E, the natural object included in one of the two natural object data Ia input to the data reception unit 2E may be, for example, a dog, and the natural object included in the other natural object data Ia may be, for example, a cat.

[0117] In this modified example, the data reception unit 2E has an interface capable of receiving input of multiple (N) natural object data Ia and one artificial object data Ib. The data reception unit 2E can input the received multiple (N) natural object data Ia and one artificial object data Ib into the generation AI unit 2F. In the generation AI unit 2F, when the learning model 11 receives multiple (N) natural object data Ia and one artificial object data Ib as input, it can generate and output composite data Ic by multiplying the multiple (N) natural object data Ia and the one artificial object data Ib together. In this modified example, the learning model 11 is a model that has been trained using, for example, multiple (N) natural object data Ia_test, one artificial object data Ib_test, and one composite data Ic_test as training data.

[0118] The product included in the synthesized data Ic is a product of multiple natural objects included in multiple (N) natural object data Ia and artificial objects included in the artificial object data Ib. When two natural object data Ia are input to the generation AI unit 2F, suppose one of the two natural object data Ia input to the generation AI unit 2F includes, for example, a dog, and the other natural object data Ia includes, for example, a cat. Furthermore, suppose the artificial object data Ib includes a vehicle wheel. In this case, the product included in the synthesized data Ic is a product of the dog, cat, and wheel.

[0119] In this modified version, the generation AI unit 2F receives multiple (N) natural object data Ia and one artificial object data Ib as input. This allows for the effective creation of more unconventional designs while presenting users with designs that are effective for idea generation.

[0120] [Modification F] Figure 33 shows a modified example of the functional block of the synthetic data generation system 2. The synthetic data generation system 2 may have interfaces that allow natural object data Ta and artificial object data Tb to be received in data receiving units 2A and 2E, for example, as shown in Figure 33. Natural object data Ta and artificial object data Tb in a common data format are input to the data receiving units 2A and 2E. The data receiving unit 2A is capable of inputting the received natural object data Ta and artificial object data Tb to the data conversion unit 2J. The data receiving unit 2E is capable of inputting the received natural object data Ta and artificial object data Tb to the data conversion unit 2K.

[0121] In this modified example, the synthetic data generation system 2 further comprises data conversion units 2J, 2K and conversion data 2L, 2M, as shown in Figure 33, for example. When natural object data Ta and artificial object data Tb are input to the data conversion unit 2J, it is possible to convert the input natural object data Ta and artificial object data Tb into a natural object data set Ia_set and artificial object data Ib using the conversion data 2L. The data conversion unit 2J is also capable of outputting the natural object data set Ia_set and artificial object data Ib obtained by the conversion to the generation AI unit 2B.

[0122] The conversion data 2L contains, for example, a set of natural object data Ia_set associated with natural object data Ta, and a set of artificial object data Ib associated with artificial object data Tb. When natural object data Ta is input to the data conversion unit 2J, it is possible to extract the set of natural object data Ia_set corresponding to the input natural object data Ta from the conversion data 2L. When artificial object data Tb is input to the data conversion unit 2J, it is possible to extract the artificial object data Ib corresponding to the input artificial object data Tb from the conversion data 2L.

[0123] The conversion data 2M contains, for example, natural object data Ia associated with natural object data Ta, and artificial object data Ib associated with artificial object data Tb. When natural object data Ta is input to the data conversion unit 2K, it is possible to extract natural object data Ia corresponding to the input natural object data Ta from the conversion data 2M. When artificial object data Tb is input to the data conversion unit 2K, it is possible to extract artificial object data Ib corresponding to the input artificial object data Tb from the conversion data 2M.

[0124] In this modified example, the natural object data Ta and artificial object data Tb received in the data reception unit 2A are converted into a natural object data set Ia_set and artificial object data Ib in the data conversion unit 2J. The converted natural object data Ia and artificial object data Ib are then input to the generation AI unit 2B. Furthermore, the natural object data Ta and artificial object data Tb received in the data reception unit 2E are converted into natural object data Ia and artificial object data Ib in the data conversion unit 2K. The converted natural object data Ia and artificial object data Ib are then input to the generation AI unit 2F. As a result, the user can obtain synthesized data Ic simply by inputting text data. Consequently, the effort required for data input by the user is reduced, while the user can be presented with designs that are effective for idea generation.

[0125] <7. Application Examples> Next, examples of applications of the synthetic data generation system 2 according to the second embodiment and modified versions E and F will be described.

[0126] Figure 34 shows an example of the internal configuration of an information processing device when the composite data generation system 2 according to the second embodiment is applied to a single information processing device. The composite data generation system 2 according to this application example includes, for example, an input unit 10, a storage unit 60, a storage unit 70, a control unit 80, and a display unit 50, as shown in Figure 34.

[0127] The input unit 10 is capable of realizing the same functions as the data reception units 2A and 2E. The input unit 10 has an interface that can accept input of natural object data set Ia_set or natural object data Ia and artificial object data Ib. The input unit 10 can input the received natural object data set Ia_set or natural object data Ia and the received artificial object data Ib to the control unit 40.

[0128] The storage unit 60 is a non-temporary tangible recording medium. The storage unit 60 is composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory. The storage unit 60 stores, for example, program 61 and learning models 11 and 13, as shown in Figure 34. Program 61 is a program that causes the control unit 80 to execute a series of procedures for realizing list display in the synthetic data generation system 2 according to this application example. Program 61 corresponds to one specific example of the "synthetic data generation program" according to one embodiment of this disclosure.

[0129] The storage unit 70 is composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory. The storage unit 70 stores reference data 72, for example, as shown in Figure 34. The reference data 72 includes selection criterion Dref4. The storage unit 70 stores reference data 71, composite data 73, and judgment result 74. The reference data 71 includes judgment criterion Dref3. The composite data 73 includes composite data Ic generated by the control unit 80. The generated composite data Ic is added to the composite data 73 each time the composite data Ic is generated by the control unit 80. The judgment result 74 includes judgment result Dx2. The generated judgment result Dx2 is added to the judgment result 74 each time the judgment result Dx2 is generated by the control unit 80. In the composite data 73 and the judgment result 74, the composite data Ic and the judgment result Dx2 are associated with each other.

[0130] The control unit 80 can execute a series of procedures to realize list display in the synthetic data generation system 2 according to this application example once program 61 is loaded into the control unit 80. The control unit 80 can realize functions similar to those of the generation AI units 2B and 2F by utilizing the learning model 11. The control unit 80 can realize functions similar to those of the filter unit 2G by utilizing the learning model 13 and reference data 71 and 72.

[0131] When the control unit 80 receives the natural object data set Ia_set and the artificial object data Ib as input, it inputs the multiple natural object data Ia included in the input natural object data set Ia_set one by one into the learning model 11, and at the same time inputs the artificial object data Ib into the learning model 11 as it inputs the natural object data Ia. When the learning model 11 receives the natural object data Ia and the artificial object data Ib as input, it can generate and output composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The learning model 11 can generate and output composite data Ic each time natural object data Ia and artificial object data Ib are input.

[0132] The control unit 80 is capable of generating an interface IF3 which includes an image listing multiple synthetic data Ic obtained by the learning model 11. Interface IF3 can, for example, enlarge and display a synthetic data Ic specified by the user. Interface IF3 can also display the synthetic data Ic selected by the user (synthetic data Ic_select) in a way that distinguishes it from the synthetic data Ic not selected by the user (synthetic data Ic_Non-select).

[0133] Interface IF3 is capable of generating a selection flag for the synthetic data Ic_select and a non-selection flag for the synthetic data Ic_Non-select. Interface IF3 is capable of storing the flag data set Flg_set, which includes one or more selection flags and one or more non-selection flags, in the storage unit 70 together with the multiple synthetic data Ic obtained by the learning model 11.

[0134] Interface IF3 is capable of storing one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70. Interface IF3 may also be capable of storing one or more feature quantities of synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70.

[0135] When the control unit 80 receives natural object data Ia and artificial object data Ib, it can input the input natural object data Ia and artificial object data Ib into the learning model 11. When the learning model 11 receives natural object data Ia and artificial object data Ib, it can generate and output composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The learning model 11 can generate and output composite data Ic each time natural object data Ia and artificial object data Ib are input.

[0136] The control unit 80 inputs the synthesized data Ic and one or more synthesized data Ic_Non-select (judgment criterion Dref3) to the learning model 13, and can obtain a judgment result Dx2 from the learning model 13 by judging the synthesized data Ic according to judgment criterion Dref3. The judgment result Dx2 is data that shows the similarity (first similarity) between the synthesized data Ic generated by the generation AI unit 2F and one or more synthesized data Ic_Non-select (judgment criterion Dref3). The judgment result Dx2 may also be data that shows the similarity (second similarity) between the feature quantities of the synthesized data Ic generated by the generation AI unit 2F and one or more feature quantities (judgment criterion Dref3) of the synthesized data Ic_Non-select.

[0137] The control unit 80 is capable of determining whether the judgment result Dx2 obtained from the learning model 13 satisfies the selection criterion Dref4. If the judgment result Dx2 satisfies the selection criterion Dref4, the control unit 80 can store the synthesized data Ic in the storage unit 70 (synthetic data 73, judgment result 74) in association with the judgment result Dx2. Each time synthesized data Ic is obtained, the control unit 80 can perform filtering based on the reference data 2H and determine whether the judgment result Dx2 satisfies the selection criterion Dref4. In this way, the control unit 80 can store multiple synthesized data Ic that satisfy the selection criterion Dref4 in the storage unit 70.

[0138] The control unit 80 is capable of outputting multiple composite data Ic that satisfy the selection criterion Dref4, stored in the storage unit 70, to the display unit 50 as a composite data list IcL. For example, the control unit 80 can generate thumbnail data for each composite data Ic that satisfies the selection criterion Dref4 read from the storage unit 70, and output the composite data list IcL, which includes the generated thumbnail data, to the display unit 50. In this way, the control unit 80 is capable of performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref4.

[0139] The control unit 80 is capable of outputting the composite data list IcL to the display unit 50 at predetermined timings. The control unit 80 is capable of outputting the composite data list IcL to the display unit 50, for example, when requested by a user of the composite generation system 2, or whenever composite data Ic that satisfies the selection criterion Dref4 is stored in the storage unit 70. The display unit 50 is capable of displaying a list of multiple composite data Ic generated by the control unit 70. When the composite data list IcL is input to the display unit 50, it is capable of generating interface IF1. The display unit 50 is configured, for example, to include a liquid crystal panel or an organic EL panel.

[0140] Next, the list display procedure in the synthetic data generation system 1 related to this application example will be described. Figure 35 shows an example of the list display procedure in the synthetic data generation system 2 related to this application example.

[0141] The control unit 80 first receives input of a natural object data set Ia_set and artificial object data Ib (step S301). The control unit 80 inputs the multiple natural object data Ia included in the received natural object data set Ia_set one by one into the learning model 11, and at the same time inputs the artificial object data Ib into the learning model 11 as the natural object data Ia is input into the learning model 11. The control unit 80 uses the learning model 11 to obtain multiple composite data Ic (step S302).

[0142] The control unit 80 displays a list of multiple synthetic data Ic obtained from the learning model 11 (step S303). The control unit 80 generates an interface IF3 which lists the multiple synthetic data Ic. Interface IF3 displays the synthetic data Ic selected by the user (synthetic data Ic_select) in a way that distinguishes it from the synthetic data Ic not selected by the user (synthetic data Ic_Non-select).

[0143] Interface IF3 generates a selection flag for the synthetic data Ic_select and a non-selection flag for the synthetic data Ic_Non-select. Interface IF3 stores the flag data set Flg_set, which includes the generated one or more selection flags and one or more non-selection flags, in the storage unit 70 along with the multiple synthetic data Ic obtained by the learning model 11. Interface IF3 stores one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70. Interface IF3 stores the features of one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70 (step S304).

[0144] The control unit 80 receives input of natural object data Ia and artificial object data Ib (step S305). The control unit 80 inputs the received natural object data Ia and artificial object data Ib into the learning model 11. The control unit 80 uses the learning model 11 to obtain synthetic data Ic (step S306). The control unit 80 inputs the synthetic data Ic and one or more synthetic data Ic_Non-select (decision criterion Dref3) into the learning model 13 and obtains the judgment result Dx2, which is the result of judging the synthetic data Ic using the judgment criterion Dref3, from the learning model 13. The control unit 80 uses the learning model 13 to obtain the judgment result Dx2 (step S307).

[0145] The control unit 80 determines whether the judgment result Dx2 obtained from the learning model 13 satisfies the selection criterion Dref4 (step S308). If the judgment result Dx2 satisfies the selection criterion Dref4, the control unit 80 stores the synthesized data Ic in the storage unit 70 in association with the judgment result Dx2 (step S309). Each time synthesized data Ic is obtained, the control unit 80 performs filtering based on the reference data 2H and determines whether the judgment result Dx2 satisfies the selection criterion Dref4. In this way, the control unit 80 stores multiple synthesized data Ic that satisfy the selection criterion Dref4 in the storage unit 70. The control unit 80 outputs the multiple synthesized data Ic stored in the storage unit 70 that satisfy the selection criterion Dref4 as a synthesized data list IcL to the display unit 50 (step S310). In this way, the list display in the synthesized data generation system 2 according to this application example is performed.

[0146] In this application example, the synthetic data generation system 2 according to the second embodiment is implemented by a single information processing device. As a result, the user can input a set of natural object data Ia_set or natural object data Ia and artificial object data Ib into a single information processing device and obtain only the design (synthetic data Ic) that is effective for idea generation.

[0147] In this application example, for example, as shown in Figure 36, a single information processing device may be configured to receive input of natural object data Ta and artificial object data Tb. In this case, the storage unit 70 further stores the natural object data 34 and the artificial object data 35.

[0148] The input unit 10 has an interface capable of receiving natural object data Ta and artificial object data Tb. The input unit 10 receives natural object data Ta and artificial object data Tb in a common data format. The input unit 10 is capable of inputting the received natural object data Ta and artificial object data Tb to the control unit 80.

[0149] When natural object data Ta and artificial object data Tb are input to the control unit 80, it is possible to extract natural object data Ia corresponding to the natural object data Ta and artificial object data Ib corresponding to the artificial object data Tb from the storage unit 70 (natural object data 34, artificial object data 35). The control unit 80 is also possible to input the natural object data Ia and artificial object data Ib extracted from the storage unit 70 into the learning model 11.

[0150] Next, a modified example of the list display procedure in the synthetic data generation system 2 according to this application example will be described. Figure 37 shows a modified example of the list display procedure in the synthetic data generation system 2 according to this application example.

[0151] The control unit 80 first receives natural object data Ta and artificial object data Tb as input (step S401). The control unit 80 uses the storage unit 70 (natural object data 34, artificial object data 35) to convert the received natural object data Ta and artificial object data Tb into a natural object data set Ia_set and artificial object data Ib (step S402). The control unit 80 inputs the multiple natural object data Ia included in the natural object data set Ia_set obtained by the conversion into the learning model 11 one by one in order, and inputs the artificial object data Ib into the learning model 11 at the same time as inputting the natural object data Ia into the learning model 11. When the learning model 11 receives natural object data Ia and artificial object data Ib as input, it generates and outputs composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The learning model 11 generates and outputs composite data Ic each time natural object data Ia and artificial object data Ib are input. The control unit 80 acquires multiple synthetic data Ic using the learning model 11 (step S403).

[0152] The control unit 80 displays a list of multiple synthetic data Ic obtained from the learning model 11 (step S404). The control unit 80 generates an interface IF3 which lists the multiple synthetic data Ic. Interface IF3 displays the synthetic data Ic selected by the user (synthetic data Ic_select) in a way that distinguishes it from the synthetic data Ic not selected by the user (synthetic data Ic_Non-select).

[0153] Interface IF3 generates a selection flag for the synthetic data Ic_select and a non-selection flag for the synthetic data Ic_Non-select. Interface IF3 stores the flag data set Flg_set, which includes the generated one or more selection flags and one or more non-selection flags, in the storage unit 70 along with the multiple synthetic data Ic obtained by the learning model 11. Interface IF3 stores one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70. Interface IF3 stores the features of one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70 (step S405).

[0154] The control unit 80 receives input of natural object data Ta and artificial object data Tb (step S406). The control unit 80 uses the storage unit 70 (natural object data 34, artificial object data 35) to convert the received natural object data Ta and artificial object data Tb into natural object data Ia and artificial object data Ib (step S407). The control unit 80 inputs the natural object data Ia and artificial object data Ib obtained by the conversion into the learning model 11. The control unit 80 uses the learning model 11 to obtain composite data Ic (step S408).

[0155] The control unit 80 inputs the synthesized data Ic and one or more synthesized data Ic_Non-select (decision criterion Dref3) into the learning model 13, and obtains a judgment result Dx2 from the learning model 13 by judging the synthesized data Ic according to the judgment criterion Dref3. The control unit 80 obtains the judgment result Dx2 using the learning model 13 (step S409).

[0156] The control unit 80 determines whether the judgment result Dx2 obtained from the learning model 13 satisfies the selection criterion Dref4 (step S410). If the judgment result Dx2 satisfies the selection criterion Dref4, the control unit 80 stores the synthesized data Ic in the storage unit 70 in association with the judgment result Dx2 (step S411). Each time synthesized data Ic is obtained, the control unit 80 performs filtering based on the reference data 2H and determines whether the judgment result Dx2 satisfies the selection criterion Dref4. In this way, the control unit 80 stores multiple synthesized data Ic that satisfy the selection criterion Dref4 in the storage unit 70. The control unit 80 outputs the multiple synthesized data Ic stored in the storage unit 70 that satisfy the selection criterion Dref4 as a synthesized data list IcL to the display unit 50 (step S412). In this way, the list display in the synthesized data generation system 2 according to this application example is performed.

[0157] In this application example, the synthetic data generation system 2 according to the second embodiment is implemented by a single information processing device. As a result, the user can input a set of natural object data Ia_set or natural object data Ia and artificial object data Ib into a single information processing device and obtain only the design (synthetic data Ic) that is effective for idea generation.

[0158] In this application example, for example, as shown in Figure 38, the composite data generation system 2 shown in Figure 34 may be implemented using terminal devices 400 and server devices 500 connected via a communication network 600. The communication network 600 may be, for example, a wired LAN such as Ethernet, a wireless LAN such as Wi-Fi, or a mobile phone line.

[0159] The terminal device 400 includes, for example, an input unit 10, a display unit 50, a storage unit 410, a control unit 420, and a communication unit 430. The communication unit 430 includes an interface capable of communicating with the server device 500 via a communication network 600. The storage unit 410 is a non-temporary tangible recording medium. The storage unit 410 is composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory. The storage unit 410 stores a program 411. The program 411 includes a series of procedures for transmitting a group of natural object data Ia_set or natural object data Ia and artificial object data Ib received by the input unit 10 to the server device 500, and for performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref4, which are included in the composite data list IcL received from the server device 500.

[0160] Once program 411 is loaded into the control unit 420, the control unit 420 can perform a series of procedures, including transmitting the natural object data group Ia_set or natural object data Ia received by the input unit 10 and the artificial object data Ib received by the input unit 10 to the server device 500, and performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref4, which are included in the composite data list IcL received from the server device 500, on the display unit 50. The display unit 50 is capable of displaying a list of multiple composite data Ic included in the composite data list IcL.

[0161] The server device 500 includes, for example, a communication unit 510, storage units 520 and 530, and a control unit 540. The communication unit 510 includes an interface capable of communicating with the terminal device 400 via a communication network 600. The storage units 520 and 530 are non-temporary tangible recording media. The storage units 520 and 530 are composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory.

[0162] The storage unit 520 stores the program 521 and the learning models 11 and 13. The program 521 includes a series of procedures for receiving the natural object data set Ia_set or natural object data Ia and the artificial object data Ib from the terminal device 400, and for generating a composite data list IcL using the natural object data set Ia_set or natural object data Ia received from the terminal device 400 and the artificial object data Ib received from the terminal device 400. The storage unit 530 stores the reference data 72. The storage unit 530 stores the reference data 71, the composite data 73, and the judgment result 74.

[0163] Once the program 521 is loaded into the control unit 540, the control unit 540 can perform a series of procedures to receive the natural object data set Ia_set or natural object data Ia and artificial object data Ib from the terminal device 400, and to generate a composite data list IcL using the natural object data set Ia_set or natural object data Ia received from the terminal device 400 and the artificial object data Ib received from the terminal device 400.

[0164] Thus, when the composite data generation system 2 shown in Figure 34 is implemented using terminal devices 400 and server devices 500 connected via a communication network 600, relatively low-spec terminal devices can be used as terminal devices 400. Furthermore, since multiple terminal devices 400 can share the server device 500, the number of terminal devices 400 can be increased at low cost.

[0165] In this application example, for example, as shown in Figure 39, the composite data generation system 2 shown in Figure 36 may be implemented using terminal devices 400 and server devices 500 connected via a communication network 600. In the server device 500, the storage unit 530 stores reference data 72, natural object data 34, and artificial object data 35. The storage unit 530 also stores reference data 71, composite data 73, and judgment results 74.

[0166] Thus, when the composite data generation system 1 shown in Figure 36 is implemented using terminal devices 400 and server devices 500 connected via a communication network 600, relatively low-spec terminal devices can be used as terminal devices 400. Furthermore, since multiple terminal devices 400 can share the server device 500, the number of terminal devices 400 can be increased at low cost.

[0167] <8. Third Embodiment> Next, a composite data generation system 3 according to a third embodiment of the present disclosure will be described. Figure 40 shows an example of the functional blocks of the composite data generation system 3 according to a third embodiment of the present disclosure. The composite data generation system 3 includes, for example, a data receiving unit 1A, a generation AI unit 1B, a filter unit 1C, reference data 1D, a storage unit 1E, a list display unit 1F, a selection flag registration unit 2C', a storage unit 2D, a data receiving unit 2E, a generation AI unit 2F, a filter unit 2G, reference data 2H, and a list display unit 2I, as shown in Figure 40.

[0168] The selection flag registration unit 2C' can assign the functionality of interface IF3 to interface IF1. Interface IF1, which has been assigned the functionality of interface IF3 (hereinafter simply referred to as "interface IF1"), can, for example, display composite data Ic selected by the user (composite data Ic_select) in a way that distinguishes it from composite data Ic not selected by the user (composite data Ic_Non-select).

[0169] Interface IF1 is capable of generating a selection flag for the synthesized data Ic_select and a non-selection flag for the synthesized data Ic_Non-select. Interface IF1 is capable of storing the flag data set Flg_set, which includes one or more selection flags and one or more non-selection flags, in the storage unit 2D together with the multiple synthesized data Ic obtained by the generation AI unit 2B.

[0170] Interface IF1 can store one or more synthetic data Ic_Non-select as the criterion Dref3 in the reference data 2H. Interface IF1 may also store the features of one or more synthetic data Ic_Non-select as the criterion Dref3 in the reference data 2H.

[0171] In this embodiment, when natural object data Ia and artificial object data Ib1 are input to the learning model 11, composite data Ic is output by multiplying the input natural object data Ia and artificial object data Ib1 together. When the composite data Ic and judgment criterion Dref3 are input to the learning model 13, judgment result Dx2 is output, which is the result of the input composite data Ic being judged by judgment criterion Dref3. Then, in the filter unit 2G, it is determined whether the judgment result Dx2 satisfies the selection criterion Dref4, and data processing is performed to display a list of multiple composite data Ics that satisfy the selection criterion Dref4. This makes it possible to eliminate unrealistic designs that are far removed from the artificial object (item) that the user wants to design, as well as safe designs that lack originality, before they are displayed in the list. As a result, designs that are effective for idea generation can be presented to the user.

[0172] In this embodiment, the synthesized data Ic is stored in the storage unit 2D in association with the judgment result Dx2. This allows the newly obtained synthesized data Ic to be displayed in a list together with multiple synthesized data Ic from the past in which the judgment result Dx2 met the selection criterion Dref4. As a result, multiple designs (synthetic data Ic) that are effective for idea generation can be presented to the user.

[0173] In this embodiment, one or more composite data Ic_Non-select or their feature quantities are stored in the reference data 2H as the judgment criterion Dref3. This allows for the exclusion of unrealistic designs that are far removed from the artifact (item) the user wants to design, as well as bland, unoriginal designs, before they are displayed in the list. As a result, the user can be presented with designs that are effective in generating ideas.

[0174] In this embodiment, the judgment result Dx2 is data indicating the first similarity or the second similarity described above, and the selection criterion Dref4 includes a threshold for the first similarity or the second similarity described above. This makes it possible to preemptively eliminate unrealistic designs that are far removed from the artifact (item) that the user wants to design, as well as bland, unoriginal designs, before they are displayed in the list. As a result, it is possible to present the user with designs that are effective in generating ideas.

[0175] <9. Modified Examples of the Third Embodiment> Next, a modified example of the synthetic data generation system 3 according to the third embodiment will be described.

[0176] Figure 41 shows a modified example of the functional block of the synthetic data generation system 3. The synthetic data generation system 3 may have interfaces that allow natural object data Ta and artificial object data Tb to be received in data receiving units 1A and 2E, as shown in Figure 41. Natural object data Ta and artificial object data Tb in a common data format are input to the data receiving units 1A and 2E. The data receiving unit 1A is capable of inputting the received natural object data Ta and artificial object data Tb to the data conversion unit 1G. The data receiving unit 2E is capable of inputting the received natural object data Ta and artificial object data Tb to the data conversion unit 2K.

[0177] In this modified example, the synthetic data generation system 3 further comprises data conversion units 1G, 2K and conversion data 1H, 2M, as shown in Figure 41, for example. When natural object data Ta and artificial object data Tb are input to the data conversion unit 1G, it is possible to convert the input natural object data Ta and artificial object data Tb into natural object data Ia and artificial object data Ib using the conversion data 1H. The data conversion unit 1G is also capable of outputting the natural object data Ia and artificial object data Ib obtained by the conversion to the generation AI unit 1B.

[0178] In this modified example, the natural object data Ta and artificial object data Tb received in the data reception unit 1A are converted into natural object data Ia and artificial object data Ib in the data conversion unit 1G. The converted natural object data Ia and artificial object data Ib are then input into the generation AI unit 2F. As a result, the user can obtain synthesized data Ic simply by inputting text data. Consequently, the effort required for data input by the user is reduced, while the user can be presented with designs that are effective for idea generation.

[0179] <10. Application Examples> Next, we will describe an example of the application of the synthetic data generation system 3 according to the third embodiment and its modified form.

[0180] Figure 42 shows an example of the internal configuration of an information processing device when the composite data generation system 3 according to the third embodiment is applied to a single information processing device. The composite data generation system 2 according to this application example includes, for example, an input unit 10, a storage unit 60, a storage unit 70, a control unit 90, and a display unit 50, as shown in Figure 42.

[0181] The input unit 10 is capable of performing the same functions as the data reception units 1A and 2E. The input unit 10 has an interface capable of receiving natural object data Ia and artificial object data Ib. The input unit 10 can input the received natural object data Ia and artificial object data Ib to the control unit 90.

[0182] The storage unit 60 stores, for example, the program 62 and the learning models 11, 12, and 13, as shown in Figure 42. The program 62 is a program that causes the control unit 90 to execute a series of procedures for realizing list display in the synthetic data generation system 3 according to this application example. The program 62 corresponds to one specific example of the "synthetic data generation program" according to one embodiment of the present disclosure. The storage unit 70 stores, for example, the reference data 31 and 72, as shown in Figure 42. The storage unit 70 stores the reference data 71, the synthetic data 73, and the judgment result 74.

[0183] The control unit 90 can execute a series of procedures to realize list display in the synthetic data generation system 3 according to this application example once program 62 is loaded into the control unit 90. The control unit 90 can realize functions similar to those of the generation AI units 1B and 2F by utilizing the learning model 11. The control unit 90 can realize functions similar to those of the filter unit 2G by utilizing the learning models 12 and 13 and reference data 71 and 72.

[0184] Next, the list display procedure in the synthetic data generation system 3 related to this application example will be described. Figure 43 shows an example of the list display procedure in the synthetic data generation system 3 related to this application example.

[0185] The control unit 90 first receives input of natural object data Ia and artificial object data Ib (step S501). The control unit 90 inputs the received natural object data Ia and artificial object data Ib into the learning model 11. The control unit 90 uses the learning model 11 to obtain composite data Ic (step S502). The control unit 90 inputs the composite data Ic and the judgment criterion Dref1 into the learning model 12. The control unit 90 uses the learning model 12 to obtain the judgment result Dx1 (step S503).

[0186] The control unit 90 determines whether the determination result Dx1 satisfies the selection criterion Dref2 (step S504). If the determination result Dx1 satisfies the selection criterion Dref2, the control unit 90 stores the synthesized data Ic in association with the determination result Dx1 in the storage unit 70 (step S505). Each time synthesized data Ic is obtained, the control unit 90 uses the learning model 12 to acquire the determination result Dx1, and if the determination result Dx1 satisfies the selection criterion Dref2, it stores the synthesized data Ic in association with the determination result Dx1 in the storage unit 70. In this way, the control unit 40 stores multiple synthesized data Ic that satisfy the selection criterion Dref2 in the storage unit 70.

[0187] The control unit 90 outputs multiple composite data Ic that satisfy the selection criterion Dref2, stored in the storage unit 70, to the display unit 50 as a composite data list IcL. The display unit 50 generates an interface IF1 that includes an image listing the multiple composite data Ic included in the composite data list IcL. As a result, the display unit 50 displays a list of multiple composite data Ic that satisfy the selection criterion Dref2 (step S506). In this way, the list display in the composite data generation system 3 according to this application example is performed.

[0188] Interface IF1 generates a selection flag for the synthetic data Ic_select and a non-selection flag for the synthetic data Ic_Non-select. Interface IF1 stores the flag data set Flg_set, which includes the generated one or more selection flags and one or more non-selection flags, in the storage unit 70 along with the multiple synthetic data Ic. Interface IF1 stores one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70. Interface IF1 stores the feature quantities of one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70 (step S507).

[0189] The control unit 90 receives input of natural object data Ia and artificial object data Ib (step S508). The control unit 90 inputs the received natural object data Ia and artificial object data Ib into the learning model 11. The control unit 90 uses the learning model 11 to obtain synthetic data Ic (step S509). The control unit 90 inputs the synthetic data Ic and one or more synthetic data Ic_Non-select (decision criterion Dref3) into the learning model 13 and obtains the judgment result Dx2, which is the result of judging the synthetic data Ic using the judgment criterion Dref3, from the learning model 13. The control unit 90 uses the learning model 13 to obtain the judgment result Dx2 (step S510).

[0190] The control unit 90 determines whether the judgment result Dx2 obtained from the learning model 13 satisfies the selection criterion Dref4 (step S511). If the judgment result Dx2 satisfies the selection criterion Dref4, the control unit 90 stores the synthesized data Ic in the storage unit 70 in association with the judgment result Dx2 (step S512). Each time synthesized data Ic is obtained, the control unit 90 performs filtering based on the reference data 2H and determines whether the judgment result Dx2 satisfies the selection criterion Dref4. In this way, the control unit 90 stores multiple synthesized data Ic that satisfy the selection criterion Dref4 in the storage unit 70. The control unit 90 outputs the multiple synthesized data Ic stored in the storage unit 70 that satisfy the selection criterion Dref4 as a synthesized data list IcL to the display unit 50 (step S513). In this way, the list display in the synthesized data generation system 3 according to this application example is performed.

[0191] In this application example, the synthetic data generation system 3 according to the third embodiment is implemented by a single information processing device. As a result, the user can input natural object data Ia and artificial object data Ib into a single information processing device and obtain only the design (synthetic data Ic) that is effective for idea generation.

[0192] In this application example, for example, as shown in Figure 44, a single information processing device may be configured to receive input of natural object data Ta and artificial object data Tb. In this case, the storage unit 70 further stores the natural object data 34 and the artificial object data 35.

[0193] The input unit 10 has an interface capable of receiving natural object data Ta and artificial object data Tb. The input unit 10 receives natural object data Ta and artificial object data Tb in a common data format. The input unit 10 can input the received natural object data Ta and artificial object data Tb to the control unit 90.

[0194] When natural object data Ta and artificial object data Tb are input to the control unit 90, it is possible to extract natural object data Ia corresponding to natural object data Ta and artificial object data Ib corresponding to artificial object data Tb from the storage unit 70 (natural object data 34, artificial object data 35). The control unit 90 is also possible to input the natural object data Ia and artificial object data Ib extracted from the storage unit 70 into the learning model 11.

[0195] Next, a modified example of the list display procedure in the synthetic data generation system 2 according to this application example will be described. Figure 45 shows a modified example of the list display procedure in the synthetic data generation system 3 according to this application example.

[0196] The control unit 90 first receives natural object data Ta and artificial object data Tb as input (step S601). The control unit 90 uses the storage unit 70 (natural object data 34, artificial object data 35) to convert the received natural object data Ta and artificial object data Tb into a natural object data set Ia_set and artificial object data Ib (step S602). The control unit 90 inputs the natural object data Ia and artificial object data Ib obtained by the conversion into the learning model 11. When the learning model 11 receives natural object data Ia and artificial object data Ib as input, it generates and outputs composite data Ic by multiplying the input natural object data Ia and artificial object data Ib together. The learning model 11 generates and outputs composite data Ic each time natural object data Ia and artificial object data Ib are input. The control unit 90 uses the learning model 11 to obtain multiple composite data Ic (step S603).

[0197] The control unit 90 inputs the synthesized data Ic and the judgment criterion Dref1 to the learning model 12. The control unit 90 uses the learning model 12 to obtain the judgment result Dx1 (step S604). The control unit 90 determines whether the judgment result Dx1 satisfies the selection criterion Dref2 (step S605). If the judgment result Dx1 satisfies the selection criterion Dref2, the control unit 90 stores the synthesized data Ic in association with the judgment result Dx1 in the storage unit 70 (step S606). Each time synthesized data Ic is obtained, the control unit 90 uses the learning model 12 to obtain the judgment result Dx1, and if the judgment result Dx1 satisfies the selection criterion Dref2, it stores the synthesized data Ic in association with the judgment result Dx1 in the storage unit 70. In this way, the control unit 40 stores multiple synthesized data Ic that satisfy the selection criterion Dref2 in the storage unit 70.

[0198] The control unit 90 outputs multiple composite data Ic that satisfy the selection criterion Dref2, stored in the storage unit 70, to the display unit 50 as a composite data list IcL. The display unit 50 generates an interface IF1 that includes an image listing the multiple composite data Ic included in the composite data list IcL. As a result, the display unit 50 displays a list of multiple composite data Ic that satisfy the selection criterion Dref2 (step S607). In this way, the list display in the composite data generation system 3 according to this application example is performed.

[0199] Interface IF1 generates a selection flag for the synthetic data Ic_select and a non-selection flag for the synthetic data Ic_Non-select. Interface IF1 stores the flag data set Flg_set, which includes the generated one or more selection flags and one or more non-selection flags, in the storage unit 70 along with the multiple synthetic data Ic. Interface IF1 stores one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70. Interface IF1 stores the feature quantities of one or more synthetic data Ic_Non-select as the decision criterion Dref3 in the storage unit 70 (step S608).

[0200] The control unit 90 receives input of natural object data Ta and artificial object data Tb (step S609). The control unit 90 uses the storage unit 70 (natural object data 34, artificial object data 35) to convert the received natural object data Ta and artificial object data Tb into natural object data Ia and artificial object data Ib (step S610). The control unit 90 inputs the natural object data Ia and artificial object data Ib obtained by the conversion into the learning model 11. The control unit 90 uses the learning model 11 to obtain the composite data Ic (step S611). The control unit 90 inputs the composite data Ic and one or more composite data Ic_Non-select (decision criterion Dref3) into the learning model 13 and obtains the judgment result Dx2 from the learning model 13 by judging the composite data Ic with the judgment criterion Dref3. The control unit 90 uses the learning model 13 to obtain the judgment result Dx2 (step S612).

[0201] The control unit 90 determines whether the judgment result Dx2 obtained from the learning model 13 satisfies the selection criterion Dref4 (step S613). If the judgment result Dx2 satisfies the selection criterion Dref4, the control unit 90 stores the synthesized data Ic in the storage unit 70 in association with the judgment result Dx2 (step S614). Each time synthesized data Ic is obtained, the control unit 90 performs filtering based on the reference data 2H and determines whether the judgment result Dx2 satisfies the selection criterion Dref4. In this way, the control unit 90 stores multiple synthesized data Ic that satisfy the selection criterion Dref4 in the storage unit 70. The control unit 90 outputs the multiple synthesized data Ic stored in the storage unit 70 that satisfy the selection criterion Dref4 as a synthesized data list IcL to the display unit 50 (step S615). In this way, the list display in the synthesized data generation system 3 according to this application example is performed.

[0202] In this application example, the synthetic data generation system 3 according to the third embodiment is implemented by a single information processing device. This allows the user to input natural object data Ta and artificial object data Tb into a single information processing device and obtain only the design (synthetic data Ic) that is effective for idea generation.

[0203] In this application example, for example, as shown in Figure 46, the composite data generation system 3 shown in Figure 42 may be implemented using terminal devices 700 and server devices 800 connected via a communication network 900. The communication network 900 may be, for example, a wired LAN such as Ethernet, a wireless LAN such as Wi-Fi, or a mobile phone line.

[0204] The terminal device 700 includes, for example, an input unit 10, a display unit 50, a storage unit 710, a control unit 720, and a communication unit 730. The communication unit 730 includes an interface capable of communicating with the server device 800 via a communication network 900. The storage unit 710 is a non-temporary tangible recording medium. The storage unit 710 is composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory. The storage unit 710 stores a program 711. The program 711 includes a series of procedures for transmitting natural object data Ia and artificial object data Ib received by the input unit 10 to the server device 800, and for performing data processing to list and display multiple composite data Ic that satisfy the selection criterion Dref4, which are included in the composite data list IcL received from the server device 800.

[0205] Once the program 711 is loaded into the control unit 720, the control unit 720 can perform a series of procedures for transmitting the natural object data Ia and artificial object data Ib received by the input unit 10 to the server device 800, and for performing data processing to display a list of multiple composite data Ic that satisfy the selection criterion Dref4, which are included in the composite data list IcL received from the server device 800, on the display unit 50. The display unit 50 is capable of displaying a list of multiple composite data Ic included in the composite data list IcL.

[0206] The server device 800 includes, for example, a communication unit 810, storage units 820 and 830, and a control unit 840. The communication unit 810 includes an interface capable of communicating with the terminal device 700 via the communication network 900. The storage units 820 and 830 are non-temporary tangible recording media. The storage units 820 and 830 are composed of, for example, non-volatile memory, such as EEPROM, flash memory, or resistive random-access memory.

[0207] The storage unit 820 stores the program 821 and the learning models 11, 12, and 13. The program 821 includes a series of procedures for receiving natural object data Ia and artificial object data Ib from the terminal device 700, and for generating a composite data list IcL using the natural object data Ia and artificial object data Ib received from the terminal device 700. The storage unit 830 stores the reference data 31 and 72. The storage unit 830 stores the reference data 71, the composite data 32 and 73, and the judgment results 33 and 74.

[0208] Once the program 821 is loaded into the control unit 840, the control unit 840 can perform a series of procedures for receiving natural object data Ia and artificial object data Ib from the terminal device 700, and for generating a composite data list IcL using the natural object data Ia and artificial object data Ib received from the terminal device 700.

[0209] Thus, when the composite data generation system 3 shown in Figure 42 is implemented using terminal devices 700 and server devices 800 connected via a communication network 900, relatively low-spec terminal devices can be used as terminal devices 700. Furthermore, since multiple terminal devices 700 can share the server device 800, the number of terminal devices 700 can be increased at low cost.

[0210] In this application example, for example, as shown in Figure 47, the composite data generation system 3 shown in Figure 44 may be implemented using terminal devices 700 and server devices 800 connected via a communication network 900. In the server device 800, the storage unit 830 stores reference data 31, 72, natural object data 34, and artificial object data 35. The storage unit 830 also stores reference data 71, composite data 32, 73, and judgment results 33, 74.

[0211] Thus, when the composite data generation system 1 shown in Figure 44 is implemented using terminal devices 700 and server devices 800 connected via a communication network 900, relatively low-spec terminal devices can be used as terminal devices 700. Furthermore, since multiple terminal devices 700 can share the server device 800, the number of terminal devices 700 can be increased at low cost.

[0212] The effects described herein are illustrative only, and the effects of this disclosure are not limited to those described herein. Therefore, other effects may be obtained with respect to this disclosure.

[0213] Furthermore, this disclosure may take the following forms: (1) A first learning model that, upon inputting one or more first data and a second data, can output composite data obtained by multiplying the input one or more first data, or one or more third data corresponding to the one or more first data, with the input second data, or a fourth data corresponding to the second data. A second learning model is provided that, upon inputting the synthesized data obtained from the first learning model and a judgment criterion, outputs a judgment result that judges the input synthesized data according to the input judgment criterion. A processing unit capable of determining whether the judgment result satisfies predetermined selection criteria and performing data processing to display a list of multiple composite data that satisfy the selection criteria. Equipped with Synthetic data generation system. (2) The aforementioned one or more first data sets are one or more 2D image data, 3D image data, or 3D model data including the first object. The second data mentioned above is one or more 2D image data, 3D image data, or 3D model data including the second object. The synthesized data is data obtained by multiplying the one or more first data and the second data, and is one or more 2D image data, 3D image data, or 3D model data that includes the product generated by the first learning model. The one or more first data, the second data, and the combined data share a common data format. (1) The synthetic data generation system described above. (3) If the one or more first data are the one 2D image data, 3D image data, or 3D model data including the first object, then the first object is a natural object. The second object is an artificial object. (2) The synthetic data generation system described above. (4) If the one or more first data are the multiple 2D image data, 3D image data, or 3D model data including the first object, then in the multiple 2D image data, 3D image data, or 3D model data, the first object is a natural object that is different from each other. The second object is an artificial object. (2) The synthetic data generation system described above. (5) The one or more first data items are one or more first character data items that represent a first object, The second data is second character data representing the second object, The aforementioned one or more third data are one or more 2D image data, 3D image data, or 3D model data including the first object, The fourth data is one or more 2D image data, 3D image data, or 3D model data including the second object. The synthesized data is data obtained by multiplying the one or more third data and the fourth data, and is one or more 2D image data, 3D image data or 3D model data including the product generated by the first learning model. The one or more third data, the fourth data, and the composite data share a common data format. (1) The synthetic data generation system described above. (6) If the one or more first data items are the one first character data item that represents the first object, then the first object is a natural object. The second object is an artificial object. (5) The composite data generation system described above. (7) If the one or more first data items are a plurality of first character data items that represent the first object, then in the plurality of first character data items, the first object is a natural object that is different from each other. The second object is an artificial object. (5) The composite data generation system described above. (8) The aforementioned determination criteria include terms that indicate whether the product included in the synthesis data is likely to be the second object. A synthetic data generation system as described in any one of (1) through (7). (9) The selection criteria include a first selection criterion indicating that the product included in the synthesis data is not similar to the second target object. (8) The synthetic data generation system described above. (10) The selection criteria include a second selection criterion indicating that the product included in the synthesis data is likely to be the second target substance. (8) The synthetic data generation system described above. (11) The processing unit is capable of storing one or more unselected data points or their features from among the multiple synthesized data output from the first learning model that were not selected by the user in the storage unit. The aforementioned determination criteria include the one or more non-selected data or their feature quantities stored in the storage unit. A synthetic data generation system as described in any one of (1) through (7). (12) The determination result is data indicating the first similarity between the synthesized data obtained from the first learning model and the unselected data, or data indicating the second similarity between the features of the synthesized data obtained from the first learning model and the features of the unselected data. The selection criteria include a threshold for the first similarity or the second similarity. (11) The synthetic data generation system described above. (13) The processing unit is capable of storing the synthesized data and the determination result in a storage unit in association with each other. A synthetic data generation system as described in any one of (1) to (12). (14) Accepting input of one or more first data and second data, By inputting the received one or more first data and the second data into the first learning model, composite data is obtained from the first learning model by multiplying the one or more first data, or one or more third data corresponding to the one or more first data, with the second data, or the fourth data corresponding to the second data. By inputting the synthesized data obtained from the first learning model and the judgment criteria into the second learning model, the judgment result obtained by judging the synthesized data according to the judgment criteria is acquired from the second learning model. The system determines whether the judgment result meets predetermined selection criteria and performs data processing to display a list of multiple composite data that meet the selection criteria. A synthetic data generation program that can be executed by a computer. (15) The 2D generated image data, the 3D generated image data, or the 3D generated model data are stored in the storage unit in association with the determination result. A synthetic data generation program as described in (14) that can be executed by a computer.

[0214] The composite data generation system 1 shown in Figures 1, 9, 13, 17, and 18, the composite data generation system 2 shown in Figures 25, 32, 33, 34, and 35, the composite data generation system 3 shown in Figures 42 and 43, the control unit 40 shown in Figures 19 and 21, the control units 120 and 420 shown in Figures 23 and 24, the control unit 80 shown in Figures 36 and 38, the control units 420 and 540 shown in Figures 40 and 41, the control unit 90 shown in Figures 44 and 46, and the control units 720 and 840 shown in Figures 48 and 49 (hereinafter referred to as "composite data generation system 1, etc.") can be implemented by a circuit including at least one semiconductor integrated circuit, such as at least one processor (e.g., a central processing unit (CPU)), at least one application-specific integrated circuit (ASIC) and / or at least one field-programmable gate array (FPGA). At least one processor can be configured to perform all or part of the various functions of the synthetic data generation system 1, etc., by reading instructions from at least one non-temporary, tangible computer-readable medium. Such a medium can take various forms, including, but is not limited to, various magnetic media such as hard disks, various optical media such as CDs or DVDs, and various semiconductor memories (i.e., semiconductor circuits) such as volatile memory or non-volatile memory. Volatile memory may include DRAM and SRAM. Non-volatile memory may include ROM and NVRAM. An ASIC is an integrated circuit (IC) specialized to perform all or part of the various functions of the synthetic data generation system 1, etc. An FPGA is an integrated circuit designed to be configurable after manufacturing to perform all or part of the various functions of the synthetic data generation system 1, etc. [Explanation of symbols]

[0215] 1,2,3…Synthetic data generation system, 1A…Data reception unit, 1B…Generation AI unit, 1C…Filter unit, 1D…Reference data, 1E…Storage unit, 1F…List display unit, 1G…Data conversion unit, 1H…Conversion data, 2A…Data reception unit, 2B…Generation AI unit, 2C,2C'…Selection flag registration unit, 2D…Storage unit, 2E…Data reception unit, 2F…Generation AI unit, 2G…Filter unit, 2H…Reference data, 2I…List display unit, 2J,2K…Data conversion unit, 2L,2M…Conversion data, 10…Input unit, 11,12…Learning model, 20…Storage unit, 21…Program M, 30...Storage unit, 31...Reference data, 32...Composite data, 33...Judgment result, 34...Natural object data, 35...Artificial object data, 40...Control unit, 50...Display unit, 60...Storage unit, 61,62...Program, 70...Storage unit, 71,72...Reference data, 73...Composite data, 74...Judgment result, 80,90...Control unit, 100,400,700...Terminal device, 110,410,710...Storage unit, 111,411,711...Program, 120,420,720...Control unit, 130,430,730...Communication unit, 200,500,800...Server device, 210,51 0,810…Communication unit, 220,520,820…Storage unit, 221,521,821…Program, 230,530,830…Storage unit, 240,540,840…Control unit, 300,600,900…Communication network, Dref1,Dref3…Judgment criteria, Dref2,Dref4…Selection criteria, Dx1,Dx2…Judgment result, Ia…Natural object data, Ia1…Natural object 2D image data, Ia2…Natural object 3D image data, Ia3…Natural object 3D model data, Ia_set…Natural object data set, Ib…Artificial object data, Ib1…Artificial object 2D image data, Ib2...3D image data of an artificial object, Ib3...3D model data of an artificial object, Ic...composite data, Ic1...2D image data of the generated product, Ic2...3D image data of the generated product, Ic3...3D model data of the generated product, IcL...composite data list, Ic_Non-select...unselected data, Ic_select...selected data, Ic_set...composite data group, IF1,IF2,IF3,IF4,IF5...interface, Flg_set...flag data group, Ta...natural object data, Ta1...natural object character data, Tb...artificial object data, Tb1...artificial object character data.

Claims

1. A first learning model that, upon input of one or more first data and a second data, can output composite data obtained by multiplying the input one or more first data, or one or more third data corresponding to the one or more first data, with the input second data, or a fourth data corresponding to the second data. A second learning model is provided that, upon inputting the synthesized data obtained from the first learning model and a judgment criterion, outputs a judgment result that judges the input synthesized data according to the input judgment criterion. A processing unit capable of determining whether the judgment result satisfies predetermined selection criteria and performing data processing to display a list of multiple composite data that satisfy the selection criteria. Equipped with Synthetic data generation system.

2. The one or more first data mentioned above are one or more 2D image data, 3D image data, or 3D model data including the first object. The second data mentioned above is one or more 2D image data, 3D image data, or 3D model data including the second object. The synthesized data is data obtained by multiplying the one or more first data and the second data, and is one or more 2D image data, 3D image data, or 3D model data that includes the product generated by the first learning model. The one or more first data, the second data, and the combined data share a common data format. The synthetic data generation system according to claim 1.

3. The one or more first data items are one or more first character data items that represent a first object, The second data is second character data that represents the second object, The one or more third data mentioned above are one or more 2D image data, 3D image data, or 3D model data including the first object. The fourth data is one or more 2D image data, 3D image data, or 3D model data including the second object. The synthesized data is data obtained by multiplying the one or more third data and the fourth data, and is one or more 2D image data, 3D image data or 3D model data including the product generated by the first learning model. The one or more third data, the fourth data, and the composite data share a common data format. The synthetic data generation system according to claim 1.

4. The processing unit is capable of storing one or more unselected data points or their feature quantities from among the multiple synthesized data output from the first learning model that were not selected by the user in the storage unit. The aforementioned determination criteria include the one or more non-selected data or their feature quantities stored in the storage unit. The synthetic data generation system according to claim 1.

5. Accepting input of one or more first data and second data, By inputting the received one or more first data and the second data into the first learning model, composite data is obtained from the first learning model by multiplying the one or more first data, or one or more third data corresponding to the one or more first data, with the second data, or the fourth data corresponding to the second data. By inputting the synthesized data obtained from the first learning model and the judgment criteria into the second learning model, the judgment result obtained by judging the synthesized data according to the judgment criteria is acquired from the second learning model. The system determines whether the judgment result meets predetermined selection criteria and performs data processing to display a list of multiple composite data that meet the selection criteria. A synthetic data generation program that can be executed by a computer.

Images