Information processing device, information processing method, and information processing program

Abstract

To provide suitable synthesis information that satisfies predetermined conditions. [Solution] The information processing device according to the present invention comprises an acquisition unit, a determination unit, and a provision unit. The acquisition unit acquires a first composite element, a second composite element, and composite information synthesized based on the first and second composite elements. The determination unit determines, based on at least the second composite element, whether the composite information satisfies predetermined conditions. The provision unit provides the composite information if the determination unit determines that the composite information satisfies predetermined conditions.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and an information processing program.

Background Art

[0002] Conventionally, a technology is known that generates a composite image from a background image, a first image, and a second image using the background image, correct answer information associated with the first image and the second image, the second image, and a composite condition associated with environment information, and generates learning data from the composite image and the correct answer information (see Patent Document 1 below).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the above conventional technology only generates a composite image from the first image, the second image, and the background image using the composite condition, and thus, for example, the above conventional technology cannot always provide suitable composite information that satisfies a predetermined condition.

[0005] The present application has been made in view of the above, and an object thereof is to provide suitable composite information that satisfies a predetermined condition.

Means for Solving the Problems

[0006] The information processing device according to the present application is characterized by comprising: a first synthesis element; a second synthesis element; an acquisition unit that acquires synthesis information synthesized based on the first and second synthesis elements; a determination unit that determines whether the synthesis information satisfies predetermined conditions based on at least the second synthesis element; and a provision unit that provides the synthesis information if the determination unit determines that the synthesis information satisfies the predetermined conditions. [Effects of the Invention]

[0007] According to one embodiment, the effect is to provide suitable synthetic information that satisfies predetermined conditions. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 shows an example of a provisioning process performed by the information processing device according to this embodiment. [Figure 2] Figure 2 shows an example of the configuration of an information processing device according to the embodiment. [Figure 3] Figure 3 shows an example of an image set storage unit according to the embodiment. [Figure 4] Figure 4 shows an example of a provided information storage unit according to the embodiment. [Figure 5] Figure 5 shows an example of a training dataset according to the embodiment. [Figure 6] Figure 6 is a flowchart showing an example of the flow of the provisioning process executed by the information processing device according to the embodiment. [Figure 7] Figure 7 shows an example of a provisioning process performed by the information processing device related to the modified example. [Figure 8] Figure 8 shows an example of a screen related to a modified example. [Figure 9] Figure 9 shows an example of a screen related to a modified example. [Figure 10] Figure 10 shows an example of a screen related to a modified example. [Figure 11] Figure 11 shows an example of a screen related to a modified example. [Figure 12] Figure 12 is a hardware configuration diagram showing an example of a computer that implements the functions of an information processing device. [Modes for carrying out the invention]

[0009] The following describes in detail, with reference to the drawings, the embodiments for implementing the information processing apparatus, information processing method, and information processing program according to the present application (hereinafter referred to as "embodiments"). Note that these embodiments do not limit the information processing apparatus, information processing method, and information processing program according to the present application. Furthermore, each embodiment can be appropriately combined as long as the processing content is not inconsistent. Also, the same parts are denoted by the same reference numerals in each of the following embodiments, and redundant descriptions are omitted.

[0010] [1. An example of a processing operation performed by an information processing device] An example of a provisioning process performed by the information processing device 100 according to the embodiment will be explained using Figure 1. Figure 1 is a diagram showing an example of a provisioning process performed by the information processing device 100 according to the embodiment.

[0011] First, the configuration of the information processing system 1 according to the embodiment will be described using Figure 1. As shown in Figure 1, the information processing system 1 includes a terminal device 10 and an information processing device 100. The terminal device 10 and the information processing device 100 are connected via a network N such as the Internet, either by wired or wireless communication. Note that the information processing system 1 shown in Figure 1 may include multiple terminal devices 10 and multiple information processing devices 100.

[0012] Terminal device 10 is an information processing device used by user U1 to access content such as web pages and application content displayed in a browser. For example, terminal device 10 may be a desktop PC (Personal Computer), a notebook PC, a tablet device, a mobile phone, or a PDA (Personal Digital Assistant).

[0013] The information processing apparatus 100 is an information processing apparatus capable of communicating with various apparatuses via a network N, and is implemented by, for example, a server apparatus, a cloud system, or the like. For example, the information processing apparatus 100 includes a first learning model. For example, the information processing apparatus 100 uses the first learning model to input a first image (an example of a first synthesis element) and a second image (an example of a second synthesis element) into the first learning model, thereby generating a synthesized image (an example of synthesis information).

[0014] For example, the first learning model is a learning model that has been trained to output a synthesized image when the first image and the second image are input into the first learning model, based on the relationship among the first image, the second image, and the synthesized image obtained by synthesizing the first image and the second image.

[0015] Further, the information processing apparatus 100 includes a second learning model. For example, the information processing apparatus 100 uses the second learning model to determine whether or not the synthesized image satisfies a predetermined condition. For example, the second learning model is a learning model that has been trained to output information related to whether or not the synthesized image satisfies a predetermined condition when the first image, the second image, and the synthesized image are input, based on the relationship among the first image, the second image, the synthesized image, and the related information of the first image, the second image, and the synthesized image. Note that the present invention is not limited to the above-described embodiment. For example, the second learning model may be a learning model that has been trained to output information related to whether or not the synthesized image satisfies a predetermined condition when the second image and the synthesized image are input, based on the relationship among the second image, the synthesized image, and the related information of the second image and the synthesized image.

[0016] Next, an example of providing processing in which the information processing apparatus 100 provides a user with information related to a synthesized image CI1 will be described with reference to FIG. 1. In the example of FIG. 1, it is assumed that a user U1 desires to evaluate whether or not the synthesized image CI1 generated from a first image FIM1 and a second image SIM1 is appropriately synthesized.

[0017] As shown in Figure 1, the information processing device 100 obtains first image information, which is information about the first image FIM1, and second image information, which is information about the second image SIM1, from the terminal device 10 or an information processing device that provides e-commerce services (including the information processing device 100 itself) (step S1). The information processing device 100 may also obtain selection operation information from the terminal device 10 to select the first image information, which is information about the first image FIM1, and selection operation information to select the second image information, which is information about the second image SIM1. For example, the first image FIM1 includes a model (person) wearing a long-sleeved shirt. The second image SIM1 includes a polo shirt.

[0018] Next, the information processing device 100 generates a composite image CI1 (step S2). For example, the information processing device 100 generates a composite image CI1 in which a model wearing a long-sleeved shirt is replaced with a polo shirt.

[0019] Then, the information processing device 100 determines, based at least on the second image SIM1, whether the composite image CI1 satisfies predetermined conditions (step S3). Specifically, the information processing device 100 determines, based at least on the second image SIM1, whether the model is properly wearing the fashion products included in the composite image CI1.

[0020] For example, the information processing device 100 inputs the first image FIM1, the second image SIM1, and the composite image CI1 into a second learning model to determine whether the composite image CI1 satisfies predetermined conditions. In the above example, the information processing device 100 has been described as determining whether the composite image CI1 satisfies predetermined conditions by inputting the first image FIM1, the second image SIM1, and the composite image CI1 into a second learning model, but it is not limited to this. For example, the information processing device 100 may determine whether the composite image CI1 satisfies predetermined conditions by inputting the second image SIM1 and the composite image CI1 into a second learning model.

[0021] To give a more specific example, the information processing device 100 determines that the composite image CI1 satisfies predetermined conditions if the model is properly wearing the fashion products included in the composite image CI1. The information processing device 100 then outputs information regarding a score indicating the degree to which the composite image CI1 satisfies the predetermined conditions, as information regarding whether or not the composite image CI1 satisfies the predetermined conditions.

[0022] On the other hand, the information processing device 100 determines that the composite image CI1 does not meet the predetermined conditions if the model is not properly wearing the fashion products included in the composite image CI1. The information processing device 100 then outputs information regarding whether the composite image CI1 meets the predetermined conditions, including a score, the area of ​​the composite image CI1 that does not meet the predetermined conditions, and at least one reason why the composite image CI1 does not meet the predetermined conditions.

[0023] Next, the information processing device 100 provides information about the composite image CI1 to the terminal device 10 (step S4). For example, if the information processing device 100 determines that the composite image CI1 satisfies predetermined conditions, it provides the terminal device 10 with information regarding a score corresponding to the determination result, which indicates the degree to which the composite image CI1 satisfies the predetermined conditions. Here, the score represents a value in the range of 0 to 1.

[0024] To give a more specific example, suppose the composite image CI1 has a score of 0.8 or higher that satisfies a predetermined condition. In this case, the information processing device 100 provides the terminal device 10 with information regarding a score of "0.9".

[0025] On the other hand, if the information processing device 100 determines that the composite image CI1 does not meet predetermined conditions, it provides the terminal device 10 with information regarding the score, the region that does not meet the predetermined conditions, and at least one reason why the composite image CI1 does not meet the predetermined conditions.

[0026] For example, if the information processing device 100 determines that the composite image CI1 does not meet predetermined conditions, it provides information regarding the score to the terminal device 10. To give a more specific example, suppose the score at which the composite image CI1 meets the predetermined conditions is 0.8 or higher. In this case, the information processing device 100 provides information regarding the score "0.5" to the terminal device 10.

[0027] As another example, if the information processing device 100 determines that the composite image CI1 does not meet predetermined conditions, it provides the terminal device 10 with information regarding the area of ​​the composite image CI1 that does not meet the predetermined conditions.

[0028] In the example shown in Figure 1, the information provided to user U1 is assumed to be provided information P1. In this case, the information processing device 100 provides the terminal device 10 with information about a rectangular region BB1 of the composite image CI1 that does not satisfy predetermined conditions, as provided information P1. Here, region BB1 is provided in a manner that it is superimposed on the composite image CI1.

[0029] As another example, if the information processing device 100 determines that the composite image CI1 does not meet predetermined conditions, it provides the terminal device 10 with information regarding the reason why the composite image CI1 does not meet the predetermined conditions. To give a more specific example, suppose the reason is that "the logo is duplicated." In this case, the information processing device 100 provides the terminal device 10 with information regarding the reason that "the logo is duplicated."

[0030] In this way, the information processing device 100 can provide useful information for the composite image CI1. For example, suppose a user who wants to purchase a second fashion product wants to find out how well the second fashion product suits them. In this case, the user might use a first image, which includes a model wearing the first fashion product (a model with similar physical characteristics to the user or the user themselves), and an image containing the second fashion product to generate a composite image that includes a model wearing the second fashion product (a model with similar physical characteristics to the user or the user themselves). The user may then use the composite image to determine whether or not the second fashion product suits them.

[0031] However, composite images often lack ground truth data (ground truth images). Therefore, conventional evaluation methods could not be used for composite images. To address this, the information processing device 100 determines, based at least on the second image SIM1, whether the composite image CI1 satisfies predetermined conditions. The information processing device 100 then provides information regarding the composite image CI1 based on the determination result. This solves the above problem. In this way, the information processing device 100 can appropriately evaluate composite images even when there is no ground truth data, such as with composite images.

[0032] [2. Configuration of the Information Processing Device] Next, an example of the functional configuration of the information processing device 100 will be described using Figure 2. Figure 2 is a diagram showing an example of the configuration of the information processing device 100 according to the embodiment. As shown in Figure 2, the information processing device 100 has a communication unit 110, a storage unit 120, and a control unit 130.

[0033] (Regarding Communications Unit 110) The communication unit 110 is implemented, for example, by a NIC (Network Interface Card). The communication unit 110 is connected to the network N by wire or wireless connection and transmits and receives information with various other devices.

[0034] (Regarding memory unit 120) The memory unit 120 is implemented by, for example, semiconductor memory elements such as RAM (Random Access Memory) and flash memory, or by storage devices such as hard disks and optical discs. The memory unit 120 also includes an image set storage unit 121, a first learning model 122, a second learning model 123, and a provided information storage unit 124.

[0035] (Regarding the image set storage unit 121) The image set storage unit 121 stores various information related to the image set. Here, Figure 3 shows an example of the image set storage unit 121 according to the embodiment. Figure 3 is a diagram showing an example of the image set storage unit 121 according to the embodiment.

[0036] In the example shown in Figure 3, the image set storage unit 121 has items such as "Image Set ID (Identifier)", "First Image", "Second Image", and "Composite Image".

[0037] The "Image Set ID" is an identifier that identifies an image set. The "First Image" is the first image information associated with the "Image Set ID". For example, the first image includes a person wearing the first fashion product.

[0038] "Second image" refers to the second image information associated with the "image set ID." For example, the second image includes the second fashion product. In the following, when there is no particular distinction between the first fashion product and the second fashion product, they may simply be referred to as "fashion product."

[0039] Here, fashion products include, for example, tops, jackets, pants, skirts, bags, shoes, fashion accessories, wallets, watches, jewelry, underwear, legwear, caps, hats, cosmetics, and other fashion items. Fashion products also include, for example, cushions, slippers, rugs, candles, furniture, lighting, clocks, wall clocks, interior decor items, photo frames, flower vases, posters, art, and other interior items. Furthermore, fashion products also include, for example, digital fashion items and digital interior items.

[0040] A "composite image" is a composite image created by combining a first image and a second image, both associated with an "image set ID." For example, a composite image may include a person wearing a second fashion product.

[0041] For example, in Figure 3, "I1," identified by the image set ID, represents the first image as "FIM1," the second image as "SIM1," and the composite image as "CI1." In the example shown in Figure 3, the first image, etc., are represented by abstract codes such as "FIM1," but the first image, etc., may also be represented by numerical values, strings, or file formats containing various information indicating the first image, etc.

[0042] (Regarding the information storage unit 124) The information provision storage unit 124 stores various types of information provision. Here, Figure 4 shows an example of the information provision storage unit 124 according to the embodiment. Figure 4 is a diagram showing an example of the information provision storage unit 124 according to the embodiment.

[0043] In the example shown in Figure 4, the provided information storage unit 124 has items such as "provided information ID," "image set ID," "score," "incompatible area," and "reason."

[0044] "Provided Information ID" is an identifier that identifies the provided information. "Image Set ID" is an identifier that identifies the image set associated with the "Provided Information ID". "Score" is information about the score output for the image set associated with the "Image Set ID".

[0045] An "incompatible area" is an area in the image set associated with the "image set ID" that does not satisfy the predetermined conditions. For example, an area that does not satisfy the predetermined conditions is displayed in the composite image in a manner in which the area that does not satisfy the predetermined conditions is surrounded by lines such as a rectangle and superimposed on the composite image. Note that the shape of the area that does not satisfy the predetermined conditions is not limited to a rectangle; it may be any shape. Also, for example, an area that does not satisfy the predetermined conditions may be displayed with the name of a part of a fashion product, such as a collar or sleeve.

[0046] The "Reason" field contains information about the reason for the output in the image set associated with the "Image Set ID". Specifically, the reason is the reason why the composite image does not match the first or second image. For example, the reason might be "The logo is duplicated." The reason may also include details of the improvement. For example, the reason might be "The logo is duplicated, but there is only one logo."

[0047] For example, in Figure 4, "P1," identified by the provided information ID, has an image set ID of "I1," a score of "SC1," a mismatched area of ​​"BB1," and a reason of "RE1."

[0048] In the example shown in Figure 4, the score and other information were represented by abstract codes such as "SC1," but the score and other information may also be represented by numerical values, strings, or file formats containing various information indicating the score.

[0049] (Regarding the control unit 130) The control unit 130 is a controller, and is implemented, for example, by a CPU (Central Processing Unit) or MPU (Micro Processing Unit) executing various programs (an example of an information processing program) stored in the memory device inside the information processing device 100 using RAM as the working area. Alternatively, the control unit 130 is a controller and can be implemented, for example, by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array).

[0050] As shown in Figure 2, the control unit 130 includes an acquisition unit 131, a generation unit 132, a determination unit 133, a provision unit 134, a reception unit 135, and a learning unit 136, and realizes or executes the information processing functions and operations described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in Figure 2, and other configurations are also possible as long as they perform the information processing described later. Also, the connection relationships of the various processing units in the control unit 130 are not limited to the connection relationships shown in Figure 2, and other connection relationships are also possible.

[0051] (Regarding acquisition unit 131) The acquisition unit 131 acquires various types of information. Specifically, the acquisition unit 131 acquires a first image, a second image, and a composite image created by combining the first and second images. For example, the acquisition unit 131 acquires first image information and second image information from a terminal device 10 or an information processing device (including the information processing device 100 itself) that provides e-commerce services. The acquisition unit 131 then stores the first image and the second image in the image set storage unit 121. The acquisition unit 131 also acquires the composite image from the image set storage unit 121.

[0052] (Regarding the generation unit 132) The generation unit 132 generates (constructs) a first learning model 122 that has been trained to output a composite image when the first image and the second image are input, based on the relationship between the first image, the second image, and the composite image. The generation unit 132 then generates a composite image by inputting the first image and the second image stored in the image set storage unit 121 into the first learning model 122. Finally, the generation unit 132 stores the generated composite image in the image set storage unit 121.

[0053] For example, the first learning model 122 is a learning model that has learned the relationship between the first image, the second image, and the composite image. To give a more specific example, the first learning model 122 is a learning model that has learned using the first image, the second image, and the composite image as ground truth data. For example, the first learning model 122 is a VTON (Virtual Try-On Network) model that generates a virtual try-on image by combining the first image, which includes a person, with the second image, which includes a fashion product.

[0054] For example, the generation unit 132 performs learning processing using methods such as backpropagation. Specifically, the generation unit 132 adjusts the values ​​of the weights (i.e., connection coefficients) that are considered when values ​​are transmitted between nodes through the learning process. In this way, the generation unit 132 learns the first learning model 122 by processing such as backpropagation to correct the parameters (connection coefficients) so that the error between the output (first output and second output) of the first learning model 122 and the correct answer (classification label) corresponding to the input is reduced. For example, the generation unit 132 generates the first learning model 122 by processing such as backpropagation to minimize a predetermined loss function. In this way, the generation unit 132 performs learning processing to learn the parameters of the first learning model 122.

[0055] Furthermore, the first learning model 122 may be a diffusion model. A diffusion model is a form of generative AI (Artificial Intelligence). For example, a diffusion model is a generative model that has learned the process of gradually adding noise to a target image to degrade it (diffusion process) and the process of gradually removing noise (denoicing) to reconstruct the image (de-diffusion process). It is assumed that the generation unit 132 has pre-generated the first learning model 122 and stored it in the storage unit 120.

[0056] Furthermore, the generation unit 132 generates (constructs) a second learning model 123 that has been trained to output information on whether the composite image satisfies predetermined conditions when the first image, second image, and composite image are input, based on the relationship between the first image, the second image, the composite image, and related information of the first image, the second image, and the composite image. It is assumed that the generation unit 132 has generated the second learning model 123 in advance and stored it in the storage unit 120. The generation process using the second learning model 123 will be described in detail in section 3 below.

[0057] (Regarding the determination unit 133) The determination unit 133 determines, based on at least the second image, whether the composite image satisfies predetermined conditions. For example, the determination unit 133 determines, based on at least the second image, whether the composite image satisfies predetermined conditions, which are the conditions for the first image and the second image to be combined. For example, suppose the synthesis condition is that the model is properly wearing the fashion products included in the composite image. In this case, the determination unit 133 determines, based on the second image, whether the model is properly wearing the fashion products included in the composite image.

[0058] For example, the determination unit 133 inputs the first image, the second image, and the composite image into the second learning model 123 to determine whether the composite image satisfies predetermined conditions. To explain with a more specific example, the determination unit 133 determines that the composite image satisfies predetermined conditions if the model is properly wearing the fashion products included in the composite image. The determination unit 133 then outputs information regarding the score as information regarding whether the composite image satisfies predetermined conditions. The determination unit 133 then stores this score information in the provided information storage unit 124.

[0059] On the other hand, the determination unit 133 determines that the composite image does not meet the predetermined conditions if the model is not properly wearing the fashion products included in the composite image. The determination unit 133 then outputs information regarding whether the composite image meets the predetermined conditions, including a score, areas that do not meet the predetermined conditions, and at least one of the reasons. The determination unit 133 then stores this information regarding the score, areas that do not meet the predetermined conditions, and at least one of the reasons in the information storage unit 124.

[0060] (Regarding Section 134) The providing unit 134 provides information corresponding to the determination result determined by the determination unit 133, and provides information about the composite image. For example, if the determination unit 133 determines that the composite image satisfies predetermined conditions, the providing unit 134 provides the terminal device 10 with information about a score corresponding to the determination result, which indicates the degree to which the composite image satisfies predetermined conditions. For example, the providing unit 134 provides the terminal device 10 with information about a score stored in the provided information storage unit 124.

[0061] To give a more specific example, suppose the composite image has a score of 0.8 or higher that satisfies a predetermined condition. In this case, the providing unit 134 provides the terminal device 10 with information regarding a score of "0.9".

[0062] On the other hand, if the determination unit 133 determines that the composite image does not meet the predetermined conditions, the providing unit 134 provides the terminal device 10 with information regarding the score, the region that does not meet the predetermined conditions, and at least one reason why the composite image does not meet the predetermined conditions.

[0063] For example, if the determination unit 133 determines that the composite image does not meet predetermined conditions, the providing unit 134 provides the terminal device 10 with information about the score stored in the providing information storage unit 124. To give a more specific example, suppose the score at which the composite image meets predetermined conditions is 0.8 or higher. In this case, the providing unit 134 provides the terminal device 10 with information about a score of "0.5".

[0064] As another example, if the determination unit 133 determines that the composite image does not meet predetermined conditions, the providing unit 134 provides the terminal device 10 with information regarding the region of the composite image that does not meet predetermined conditions. For example, the providing unit 134 provides the terminal device 10 with information regarding the region that does not meet predetermined conditions, which is stored in the provided information storage unit 124.

[0065] To give a more specific example, the providing unit 134 provides the terminal device 10 with information about a rectangular region BB1 in the composite image that does not satisfy predetermined conditions. Here, region BB1 is provided in a manner that it is superimposed on the composite image.

[0066] As another example, if the determination unit 133 determines that the composite image does not meet predetermined conditions, the providing unit 134 provides the terminal device 10 with information regarding the reason. For example, the providing unit 134 provides the terminal device 10 with information regarding the reason stored in the providing information storage unit 124.

[0067] To give a more specific example, let's say the reason is "the logo is duplicated." In this case, the providing unit 134 provides the terminal device 10 with information regarding the reason, "the logo is duplicated."

[0068] (Regarding reception desk 135) The reception unit 135 receives various types of information. For example, the reception unit 135 receives annotation-related information from the terminal device 10. The reception processing performed by the reception unit 135 is described in detail in section 5-3 below.

[0069] (Regarding Learning Section 136) The learning unit 136 trains the first learning model 122, which has been trained to output a composite image when the first image and the second image are input, to learn the relationship between the first image, the second image, the composite image, and the judgment result. The learning process performed by the learning unit 136 will be described in detail in 5-5 below.

[0070] [3. Generation process of the second learning model 123] Next, the generation process for generating the second learning model 123 will be described. For example, the generation unit 132 generates a second learning model 123 that has been trained to output information on whether or not the composite image satisfies predetermined conditions when the first image, the second image, and the composite image are input, based on the relationship between the first image, the second image, the composite image, and related information.

[0071] For example, the generation unit 132 trains the second learning model 123 to calculate a score corresponding to the appropriateness indicated by the relevant information when the first image, the second image, and the composite image are input. For example, the generation unit 132 trains the second learning model 123 to output a higher score if the appearance of the fashion product in the composite image is appropriate in comparison with the second image. Conversely, the generation unit 132 trains the second learning model 123 to output a lower score if the appearance of the fashion product in the composite image is not appropriate in comparison with the second image.

[0072] Furthermore, if the score is an inappropriateness score, the generation unit 132 trains the second learning model 123 to output a higher score when the appearance of the fashion product in the composite image is not appropriate in comparison to the second image. Also, if the appearance of the fashion product in the composite image is appropriate in comparison to the second image, the generation unit 132 trains the second learning model 123 to output a lower score.

[0073] Furthermore, the generation unit 132 is not limited to the above example, and may train the second learning model 123 to lower (or raise) the score the more inappropriate parts or regions there are. In addition, the generation unit 132 may also input relevant information to the second learning model 123 during training, thereby allowing the second learning model 123 to learn the reasons for the inappropriateness.

[0074] In the following section, we will describe the training dataset for generating the second learning model 123 using Figure 5. Figure 5 is a diagram showing an example of a training dataset according to the embodiment. The training dataset is data that associates image IM1, image IM2, composite image CI1, and related information.

[0075] In the example in Figure 5, image IM1 includes the second fashion product. Image IM2 includes a person wearing the first fashion product. Composite image CI1 includes a person wearing the second fashion product. For example, comparing image IM1 with composite image CI1, composite image CI1 shows a round neck in the second fashion product, whereas the second fashion product in IM1 has a V-neck. Also, composite image CI1 does not adequately reflect the striped pattern on the right side. Furthermore, the printed text on the front of composite image CI1 is distorted.

[0076] For example, comparing image IM1 with composite image CI1, the areas in image IM1 where the second fashion product is not properly reflected are designated as areas RC11, RC12, and RC13. In the example in Figure 5, area RC11 is the area around the neck of the second fashion product. Area RC12 is the area of ​​the right side of the second fashion product. Area RC13 is the area on the front of the second fashion product that includes the printed text.

[0077] Furthermore, the regions of composite image CI1 corresponding to areas where the second fashion product is not properly reflected are designated as regions RC31, RC32, and RC33. Additionally, the regions of image IM2 corresponding to areas where the second fashion product is not properly reflected are designated as regions RC21, RC22, and RC23.

[0078] For example, the related information includes information linking regions RC11, RC21, and RC31. It also includes information linking regions RC12, RC22, and RC32. Furthermore, it includes information linking regions RC13, RC23, and RC33. The related information may also include information regarding the reason why regions RC11, RC21, and RC31 are "round necks, not V-necks." Thus, the related information is information linking regions in each image where the second fashion product is not properly reflected with the reason. The related information also includes information regarding the appropriateness score (or inappropriateness score) indicating whether the second fashion product is properly reflected in the composite image CI1.

[0079] The related information is not limited to the above example and may also be information about annotations input by the annotator, as detailed in 5-3 below. Furthermore, in the above embodiment, the generation unit 132 was described as generating a second learning model 123 that has been trained to output information on whether or not the composite image satisfies predetermined conditions when the first image, the second image, and the composite image are input, based on the relationship between the first image, the second image, the composite image, and the related information, but the embodiment is not limited to this. For example, the generation unit 132 may generate a second learning model 123 that has been trained to output information on whether or not the composite image satisfies predetermined conditions when the second image and the composite image are input, based on the relationship between the second image, the composite image, and the related information.

[0080] [4. Processing Procedure] Next, the procedure for the provisioning process executed by the information processing device 100 according to the embodiment will be explained using Figure 6. Figure 6 is a flowchart showing an example of the flow of the provisioning process executed by the information processing device 100 according to the embodiment.

[0081] As shown in Figure 6, the acquisition unit 131 acquires the first composite element, the second composite element, and the composite information (step S101). Specifically, if the acquisition unit 131 has not acquired the first composite element, the second composite element, and the composite information (step S101; No), it waits until it acquires the first composite element, the second composite element, and the composite information.

[0082] On the other hand, if the acquisition unit 131 has acquired the first composite element, the second composite element, and the composite information (step S101; Yes), the determination unit 133 determines, based at least on the second composite element, whether the composite information satisfies predetermined conditions (step S102).

[0083] Specifically, if the determination unit 133 determines that the composite information satisfies predetermined conditions (step S102; Yes), the providing unit 134 provides first information corresponding to the determination result (step S103). For example, as first information, the providing unit 134 provides information regarding the score corresponding to the determination result.

[0084] On the other hand, if the determination unit 133 determines that the composite information does not meet predetermined conditions (step S102; No), the providing unit 134 provides second information corresponding to the determination result (step S104). For example, as second information, the providing unit 134 provides information relating to at least one of the following: a score corresponding to the determination result, areas that do not meet predetermined conditions, and reasons.

[0085] [5. Variations] The information processing device 100 described above may be implemented in various other forms besides those described above. Therefore, other embodiments of the information processing device 100 will be described below.

[0086] [5-1. Combination elements and combination information] In the above embodiment, an example was given where the composite element is an image, but the invention is not limited to this. For example, the composite element may be a video or a 3D model. Similarly, an example was given where the composite information is an image, such as a composite image, but the invention is not limited to this. For example, the composite information may be a video or a 3D model.

[0087] Furthermore, although the above embodiments have described examples where the image includes a fashion product or a model wearing a fashion product, the embodiments are not limited to these examples. The image may also include, for example, a natural landscape such as a mountain.

[0088] [5-2. Separation of Information Processing Device 100] Furthermore, although the explanation described an example in which the information processing device 100 has a first learning model 122, it is not limited to this. For example, the information processing device 100 does not have to have a first learning model 122. In this case, another external server different from the information processing device 100 may have the first learning model 122.

[0089] In this case, the external server receives the first and second composite elements from the information processing device 100. Subsequently, the external server generates composite information by inputting the first and second composite elements into the first learning model 122 using the first learning model 122. The external server may then provide the generated composite information to the information processing device 100. For example, the external server may implement the above information processing as an API (Application Programming Interface) function.

[0090] [5-3. Annotation] Furthermore, the reception unit 135 may also receive annotation-related information from the terminal device 10 used by the annotator. The annotation-related information referred to here is information related to the evaluation of the synthesized information. For example, the annotation-related information is the label assigned to the synthesized information by the annotator.

[0091] Below, we will explain an example of how the information processing device 100 provides information about an annotation tool, using Figure 7. Figure 7 is a diagram showing an example of the processing performed by the information processing device 100 according to a modified example.

[0092] As shown in Figure 7, the information processing device 100 provides information about the annotation tool (step S21). The annotation tool referred to here is a tool that causes an annotator to annotate the composite image.

[0093] Here, we will explain information about the annotation tool using Figures 8 to 10. In the example in Figure 8, screen C1 of the annotation tool is shown. Figure 8 is a diagram showing an example of screen C1 related to a modified example.

[0094] Screen C1 includes the first image IM11, the second image IM12, the reference image (third image IM13), the composite image CI11, and the annotation items AT1 to AT7. The first image IM11 includes the first model wearing the first fashion product. In the example in Figure 8, the first image IM11 includes the first model wearing the first T-shirt.

[0095] The second image IM12 includes the second fashion product. In the example in Figure 8, the second image IM12 includes the second T-shirt. The third image IM13 includes the second model wearing the second fashion product, and is different from the first model included in the first image IM11. In the example in Figure 8, the third image IM13 includes the second model wearing the second T-shirt.

[0096] The composite image CI11 is an image created by combining the first image IM11 and the second image IM12. For example, composite image CI11 includes the first model wearing the second fashion product. In the example in Figure 8, composite image CI11 includes the first model wearing the second T-shirt.

[0097] Item AT1 is for confirming the color of the second fashion product. Item AT2 is for confirming the pattern of the second fashion product. Item AT3 is for confirming the design of the second fashion product.

[0098] Item AT4 is for checking the silhouette of the secondary fashion product. Item AT5 is for checking the wrinkles of the secondary fashion product. Item AT6 is for checking the placement of the secondary fashion product. Item AT7 is for checking the overall appearance of the secondary fashion product. Note that the items are not limited to these, and there may also be an item for checking the size of the secondary fashion product.

[0099] In the example in Figure 8, it is assumed that item AT1 is pressed by the annotator to check the color of the second fashion product. In this case, on screen C1, the second image IM12 is displayed with a color bar superimposed. In the example in Figure 8, the color bar displays AI11, which is information about the color sample of the currently selected part in the second image IM12 (the color sample of the currently selected part in the second fashion product). The annotator can check the color of the second fashion product contained in the second image IM12 by comparing the second image IM12 with the information about the color sample AI11.

[0100] Furthermore, the third image IM13 is displayed with a color bar superimposed on it. In the example in Figure 8, the color bar displays AI12, which is information about the color sample of the currently selected part in the third image IM13 (the color sample of the currently selected part in the second fashion product). The annotator can confirm the color of the second fashion product contained in the third image IM13 by comparing the third image IM13 with the information about the color sample AI12.

[0101] Furthermore, the composite image CI11 is displayed with a color bar superimposed on it. In the example in Figure 8, the color bar displays CR11, which is information about the color sample of the currently selected part in the composite image CI11 (the color sample of the currently selected part in the second fashion product). The annotator can confirm the color of the second fashion product included in the composite image CI11 by comparing the composite image CI11 with the color sample information CR11. In other words, the annotator can confirm whether the color of the second fashion product included in the composite image CI11 is appropriately reproduced by referring to the color confirmed by comparing the second image IM12 with the color sample information AI11, the color confirmed by comparing the third image IM13 with the color sample information AI12, and the color confirmed by comparing the composite image CI11 with the color sample information CR11. Colors can appear different even if they are the same color due to the influence of surrounding colors. For example, in the second image IM12, when the color of the second fashion product is navy blue and the background color is white, and in the composite image CI11, when the color of the second fashion product is navy blue and the background color is black or the color of the fashion product it is combined with is a dark color, the color of the second fashion product may appear slightly different even though the color is the same. Displaying a color bar can solve this problem. In addition, the annotator can specify the color of the second fashion product included in the second image IM12, the color of the second fashion product included in the third image IM13, and the color of the second fashion product included in the composite image CI11 by manipulating the color sample information AI11, the color confirmed by comparing the third image IM13 with the color sample information AI12, and the color confirmed by comparing the composite image CI11 with the color sample information CR11.This allows the annotator to annotate that the color of the second fashion product included in the composite image CI11 is properly reproduced if the color confirmed by comparing the second image IM12 with the color sample information AI11 (or the color confirmed by comparing the third image IM13 with the color sample information AI12) is the same as the color confirmed by comparing the composite image CI11 with the color sample information CR11. Conversely, if the color confirmed by comparing the second image IM12 with the color sample information AI11 (or the color confirmed by comparing the third image IM13 with the color sample information AI12) is different from the color confirmed by comparing the composite image CI11 with the color sample information CR11, the annotator can annotate that the color of the second fashion product included in the composite image CI11 is not properly reproduced. Furthermore, the annotator may, by referring to the colors confirmed by comparing the second image IM12 with the color sample information AI11, the colors confirmed by comparing the third image IM13 with the color sample information AI12, and the colors confirmed by comparing the composite image CI11 with the color sample information CR11, input in a separate input field whether or not the colors of the second fashion product included in the composite image CI11 are appropriately reproduced. In other words, the annotator can provide a score indicating appropriateness (or inappropriateness) and information on at least one area where the color is not appropriately reproduced and the reason (including improvements).

[0102] Next, we will explain an example of checking the pattern of the second fashion product using Figure 9. In the example in Figure 9, screen C2 of the annotation tool is shown. Figure 9 is a diagram showing an example of screen C2 related to a modified example.

[0103] Screen C2 includes the first image IM21, the second image IM22, the third image IM23, the composite image CI21, and annotation items AT1 to AT7. In the example in Figure 9, item AT2 is assumed to have been pressed by the annotator to confirm the pattern of the second fashion product.

[0104] In this case, the second image IM22, in a form where the pattern is emphasized, is displayed on screen C2. In the example in Figure 9, the pattern contained in region AI21 is displayed in a form that can be manipulated by the annotator. For example, these manipulations include selecting, moving, scaling, correcting pattern distortion, and correcting pattern tilt.

[0105] Furthermore, a third image IM23 is displayed in a form in which the pattern is emphasized. In the example in Figure 9, the pattern contained in region AI22 is displayed in a form that can be manipulated by the annotator.

[0106] Furthermore, a composite image CI21 with the pattern emphasized is displayed. In the example in Figure 9, the pattern contained in region CR21 is displayed in a manner that can be manipulated by the annotator. In addition, the pattern contained in region AI21 and the pattern contained in region AI22 can be moved as reference information within region CR21. In this way, the annotator can check whether the pattern of the second fashion product contained in the composite image CI31 is appropriately reproduced by referring to the reference information. The annotator can also manipulate the pattern contained in region CR21 (correcting pattern distortion, correcting pattern tilt, etc.). As a result, if the pattern contained in region AI21 (or region AI22) and the pattern contained in region CR21 are the same, the annotator can annotate that the pattern of the second fashion product contained in the composite image CI21 is appropriately reproduced. If the pattern contained in region AI21 (or region AI22) and the pattern contained in region CR21 are different, the annotator can annotate that the pattern of the second fashion product contained in the composite image CI21 is not appropriately reproduced. Furthermore, the annotator may, by referring to the reference information, input in a separate field whether or not the pattern of the second fashion product included in the composite image CI21 is appropriately reproduced. That is, the annotator can provide a score indicating appropriateness (or inappropriateness) and information on at least one area where the pattern is not appropriately reproduced and the reason (including improvements). Note that the reference information is not limited to the above modifications, and a reference image containing only the pattern may be used as the reference information.

[0107] Next, we will explain an example of checking wrinkles in a second fashion product using Figure 10. In the example in Figure 10, screen C3 of the annotation tool is shown. Figure 10 is a diagram showing an example of screen C3 related to a modified example.

[0108] Screen C3 includes the first image IM31, the second image IM32, the third image IM33, the composite image CI31, and annotation items AT1 to AT7. In the example in Figure 10, item AT5 is pressed by the annotator to check for wrinkles in the second fashion product.

[0109] In this case, the first image IM31 in a form where wrinkles are emphasized is displayed on screen C3. In the example in Figure 10, the wrinkles contained in region AI31 are displayed in a form that can be manipulated by the annotator.

[0110] Furthermore, a composite image CI31 with emphasized wrinkles is displayed. In the example in Figure 10, the wrinkles contained in region CR31 are displayed in a manner that can be manipulated by the annotator. In this way, the annotator can refer to region AI31 to confirm whether the wrinkles of the second fashion product contained in the composite image CI31 are appropriately reproduced. The annotator can also manipulate the wrinkles contained in region CR31 (correcting wrinkle distortion, correcting wrinkle tilt, etc.). As a result, if the wrinkles contained in region AI31 and the wrinkles contained in region CR31 are the same, the annotator can annotate that the wrinkles of the second fashion product contained in the composite image CI31 are appropriately reproduced. If the wrinkles contained in region AI31 and the wrinkles contained in region CR31 are different, the annotator can annotate that the wrinkles of the second fashion product contained in the composite image CI31 are not appropriately reproduced. Furthermore, the annotator may, by referring to region AI31, input in a separate input field whether or not the wrinkles of the second fashion product included in the synthesized image CI31 are appropriately reproduced. In other words, the annotator can provide a score indicating appropriateness (or inappropriateness) and information on at least one region where the wrinkles are not appropriately reproduced and the reason (including improvements).

[0111] Next, as shown in Figure 7, the information processing device 100 receives annotation-related information from the terminal device 10 (step S22). For example, the information processing device 100 receives information regarding the color, pattern, or wrinkles confirmed in the composite image (including information regarding a score indicating appropriateness (or inappropriateness)) from the terminal device 10 as annotation-related information.

[0112] Here, the information processing device 100 may, for example, train the first learning model 122 on the relationship between the first image, the second image, the composite image, and the annotation information. Alternatively, the information processing device 100 may, for example, train the first learning model 122 on the relationship between the first image, the second image, the composite image, and the appropriateness (or inappropriateness) of the composite image (also referred to as information on a score indicating the appropriateness (or inappropriateness) of the composite image, information on inappropriate regions and / or reasons for the composite image, or information on improvements to the composite image). In this case, for example, the first and second images are specified by an annotator or a user. At this time, the information processing device 100 generates a composite image by inputting the first and second images specified by the annotator or a user into the first learning model 122 (step S23). At this time, the information processing device 100 may generate a composite image by inputting the first and second images specified by the annotator or a user, and the information on the appropriateness of the composite image into the first learning model 122.

[0113] Next, the information processing device 100 provides the composite image information to the terminal device 10 (step S23). This allows the annotator to refer to the composite image and determine whether the quality of the composite image has improved. The user can then receive the composite image with improved quality. In this way, the information processing device 100 can improve the quality of the composite image.

[0114] Furthermore, the information processing device 100 may, for example, train the second learning model 123 to learn the relationships between the first image, the second image, the composite image, and related information such as annotations. This allows the information processing device 100 to accurately output information regarding whether or not the composite image satisfies predetermined conditions using the second learning model 123.

[0115] In the above modified example, we explained the case where the annotator selects one annotation item, but this is not the only example. For example, multiple annotation items may be selected. An example in which multiple annotation items are selected will be explained using Figure 11. In the example in Figure 11, the annotation tool screen C4 is shown. Figure 11 is a diagram showing an example of screen C4 related to the modified example.

[0116] Screen C4 includes the first image IM41, the second image IM42, the third image IM43, the composite image CI41, and annotation items AT1 to AT7. In the example in Figure 11, items AT2 and AT5 are assumed to have been pressed by the annotator to check the pattern and wrinkles of the second fashion product.

[0117] In this case, the first image IM41 in a manner in which wrinkles are emphasized is displayed on screen C4. In the example in Figure 11, the wrinkles contained in region AI41 are displayed in a manner that can be manipulated by the annotator.

[0118] Furthermore, a second image IM42 in which the pattern is emphasized is displayed. In the example in Figure 11, the pattern contained in region AI42 is displayed in a manner that can be manipulated by the annotator.

[0119] Furthermore, a third image IM43 is displayed in a form in which the pattern is emphasized. In the example in Figure 11, the pattern contained in region AI43 is displayed in a form that can be manipulated by the annotator.

[0120] Furthermore, a composite image CI41 is displayed in a manner that emphasizes the patterns and wrinkles. In the example in Figure 11, the patterns and wrinkles contained in region CR41 are displayed in a manner that can be manipulated by the annotator. In addition, the patterns contained in region AI21 and region AI22 can be moved within region CR41 as reference information. In this way, the annotator can check whether the patterns and wrinkles of the second fashion product contained in the composite image CI41 are appropriately reproduced by referring to regions AI41, AI42, and AI43.

[0121] Furthermore, an image containing a model with similar physical characteristics to the first model in the first image may be displayed on the screen as the third image (this minimizes the possibility of the same object appearing different due to differences in the models' physical characteristics). There may also be multiple third images. Additionally, the annotation items are not limited to the above modifications; they may be displayed in a selectable manner, allowing users to choose the model's skin color, body type, hairstyle, posture, orientation, etc. In this case, the annotator will verify each item. The annotation items may also include an option to select whether the fashion items worn by the model are layered.

[0122] Furthermore, the information processing device 100 may provide the annotator with information such as typical reasons for failure and typical areas that do not meet predetermined conditions, as examples of information for annotation.

[0123] [5-4. Rule-based] Furthermore, the determination unit 133 may determine whether the composite information satisfies predetermined conditions based on a predetermined evaluation policy. The predetermined evaluation policy is information used to evaluate the composite information. For example, the predetermined evaluation policy is set in advance by an administrator who manages the information processing device 100. In this case, the storage unit 120 stores information related to the predetermined evaluation policy.

[0124] To give a more specific example, let's assume that the composite elements and composite information are images. The first image includes a model wearing the first fashion product. The second image includes the second fashion product. The composite image also includes a model wearing the second fashion product. In this case, the predetermined evaluation policy is information for evaluating whether the model is wearing the second fashion product included in the composite image appropriately. For example, the predetermined evaluation policy is information for evaluating the color, pattern, design, silhouette, wrinkles, placement of logos, etc., of the second fashion product, as well as the overall appearance of the second fashion product.

[0125] Furthermore, the predetermined evaluation policy may be determined based on attribute information relating to the user's attributes and the situations in which the user wants to generate composite information. For example, attribute information may include demographic attributes and psychographic attributes. For example, demographic attributes are demographic attributes. More specifically, demographic attributes include age, gender, occupation, place of residence, annual income, family structure, etc. For example, psychographic attributes are psychological attributes. More specifically, psychographic attributes include lifestyle, values, interests, preferences for fashion products or brands of fashion products, etc. Based on this, the determination unit 133 can determine whether the composite information meets predetermined conditions by considering the aspects that the user values ​​in the composite image and the aspects that the user values ​​in the fashion products.

[0126] For example, the predetermined evaluation policy may be determined for each element such as color, pattern, design, silhouette, wrinkles, arrangement, and overall appearance, based on various trends of the user or shop (brand, product, etc.) and the content of the service. This allows the determination unit 133 to determine whether the composite information meets the predetermined conditions under conditions favorable to the user.

[0127] [5-5. First Learning Model 122] Furthermore, the learning unit 136 may train the first learning model 122 on the relationships between the first composite element, the second composite element, the composite information, and information regarding the judgment result (information regarding annotations and information regarding whether the composite information output by the second learning model 123 satisfies predetermined conditions). For example, the learning unit 136 may train the first composite element, the second composite element, the composite information, and at least one of the following: a score for the composite information satisfying predetermined conditions, a region that does not satisfy the predetermined conditions, and a reason. This allows the learning unit 136 to update the first learning model 122 so that it can output more accurate composite information by inputting the first composite element, the second composite element, the composite information, and information regarding the judgment result. The learning unit 136 may also update the first learning model 122 so that it can output more accurate composite information by inputting the second composite element, the composite information, and information regarding the judgment result. This allows the learning unit 136 to improve the quality of the composite information. Furthermore, if the first learning model 122 is a diffusion model, the learning unit 136 may input a first composite element, a second composite element, composite information, and information regarding the judgment result, and output more accurate composite information (composite information that satisfies predetermined conditions). Furthermore, if the first learning model 122 is a diffusion model, the learning unit 136 may input a second composite element, composite information, and information regarding the judgment result, and output more accurate composite information (composite information that satisfies predetermined conditions).

[0128] Furthermore, the first learning model 122 may be a learning model that, when given at least a first image including the user and a second image including the fashion products the user wishes to wear, outputs (generates) a composite image in which the fashion products included in the second image are worn in a manner that suits the user included in the first image.

[0129] For example, suppose that multiple annotators have been pre-instructed to evaluate target images. The target images are images showing models wearing fashion products. In this case, suppose that images and their evaluation values ​​(evaluation values ​​that assess whether the fashion product suits the model wearing it) are pre-associated and stored in the storage unit 120. In such a case, the generation unit 132 may generate a first learning model 122 that has been trained with images whose evaluation values ​​are above a predetermined threshold as ground truth data, or a first learning model 122 that has been trained with the relationship between images whose evaluation values ​​are above a predetermined threshold and such evaluation values. As a result, the generation unit 132 can generate images that satisfy evaluation criteria such as whether the fashion product suits the user.

[0130] Furthermore, the learning unit 136 may train the first learning model 122 with images that were evaluated as suitable from a male perspective. Alternatively, the learning unit 136 may train the first learning model 122 with images that were evaluated as suitable from a female perspective. The first learning model 122 trained with images that were evaluated as suitable from a male perspective and the first learning model 122 trained with images that were evaluated as suitable from a female perspective may be switched between depending on the user or at the request of the user.

[0131] [5-6. Second Learning Model 123] Furthermore, the second learning model 123 may be an evaluation model for generating the training dataset. For example, the evaluation model may be a large-scale language model such as a transformer-based model.

[0132] Furthermore, the second learning model 123 may be a learning model that has been trained on evaluation information evaluated by the annotator, such as absolute or relative evaluations indicating whether the fashion products included in the image suit the model wearing them.

[0133] Furthermore, the second learning model 123 may be a learning model that has been trained considering the order in which images are presented to the annotator to allow the annotator to evaluate the evaluation information evaluated by the annotator, and which include images of models wearing fashion products.

[0134] Furthermore, the second learning model 123 may be a learning model generated based on a combination of a learning model that was trained considering the presentation order of the images presented to the annotator in order to evaluate the evaluation information evaluated by the annotator, and a learning model that was trained without considering the presentation order of the images in order to evaluate the evaluation information evaluated by the annotator.

[0135] [5-7. Acquisition Process] Furthermore, the acquisition unit 131 may acquire a first composite element in which the region where the second composite element is synthesized is masked (a first composite element in which noise is added to the region where the second composite element is synthesized). For example, suppose the first composite element is a first image including a model wearing a first fashion product. In this case, the acquisition unit 131 may acquire a first image in which the region corresponding to the first fashion product is masked. Specifically, the acquisition unit 131 may acquire a first image in which the region corresponding to the first fashion product, which is in the same category as the second fashion product included as the second composite element, is masked. The learning unit 136 may then input the first composite element in which the region where the second composite element is synthesized is masked, and the second composite element, to the first learning model 122 and output the synthesis information.

[0136] Furthermore, the masked first composite element may be used in the training process of the first learning model 122. For example, the learning unit 136 may learn the relationship between the first composite element, the second composite element, the composite information, and the masked first composite element. This allows the learning unit 136 to update the first learning model 122 so that it can output composite information with higher accuracy. This allows the learning unit 136 to improve the quality of the composite information.

[0137] Furthermore, the acquisition unit 131 may acquire composite information in which the regions of the composite information that do not satisfy predetermined conditions are masked (composite information in which noise is added to the regions of the composite information that do not satisfy predetermined conditions). For example, suppose the composite information is a composite image including a model wearing the second fashion product. Also, suppose the region that does not satisfy predetermined conditions is a region including the pattern of the second fashion product. In this case, the acquisition unit 131 may acquire a composite image in which the region including the pattern of the second fashion product is masked. The learning unit 136 may then input the first composite element in which the region of the first composite element that synthesizes the second composite element is masked, the second composite element, the composite information in which the regions of the composite information that do not satisfy predetermined conditions are masked, and information regarding the judgment result (information regarding whether the composite information output by the second learning model 123 satisfies predetermined conditions, including information regarding the score, region, and reason) to the first learning model 122 and output more accurate composite information (composite information that satisfies predetermined conditions). Alternatively, the learning unit 136, the second synthesis element, the synthesized information in which regions that do not satisfy predetermined conditions are masked, and information regarding the judgment result (information regarding whether the synthesized information output by the second learning model 123 satisfies predetermined conditions) may be input to the first learning model 122 to output more accurate synthesized information (synthesized information that satisfies predetermined conditions).

[0138] Furthermore, the acquisition unit 131 may acquire a second composite element in which the region of the second composite element that does not satisfy a predetermined condition is masked. For example, suppose the second composite element is a second image including a second fashion product. Also, suppose the region that does not satisfy the predetermined condition is the region including the logo of the second fashion product. In this case, the acquisition unit 131 may acquire a second image in which the region including the logo of the second fashion product is masked.

[0139] Furthermore, the masked second composite element may be used in the training process of the second learning model 123. For example, the learning unit 136 may further learn the relationships between the first composite element, the second composite element, the composite information, the related information, and the masked second composite element. This allows the learning unit 136 to update the second learning model 123 so that it can more accurately determine whether the composite information satisfies predetermined conditions.

[0140] [5-8. Extracted Information] Furthermore, the acquisition unit 131 may acquire the first extracted information extracted from the first composite element. The determination unit 133 may then use the first extracted information to determine whether the composite information satisfies predetermined conditions.

[0141] For example, suppose the first composite element is a first image including a model wearing the first fashion product. Also, suppose the first fashion product has wrinkles. Also, suppose the second composite element is a second image including the second fashion product. Also, suppose the composite information is a composite image including a model wearing the second fashion product. In this case, the acquisition unit 131 acquires information about wrinkles extracted from the first image as the first extracted information. Note that the extraction process for extracting information about wrinkles from the first image can be realized by conventional techniques such as image analysis.

[0142] Furthermore, the determination unit 133 may use information about wrinkles to determine whether the composite image satisfies predetermined conditions. For example, the determination unit 133 may determine, based on at least the second image and the information about wrinkles, whether the model is properly wearing the fashion product included in the composite image. This allows the determination unit 133 to determine with greater accuracy whether the composite information satisfies predetermined conditions.

[0143] Furthermore, the acquisition unit 131 may acquire second extraction information extracted from the second composite element. The determination unit 133 may then use the second extraction information to determine whether the composite information satisfies predetermined conditions.

[0144] For example, suppose the first composite element is a first image including a model wearing the first fashion product. Also, suppose the second composite element is a second image including the second fashion product. Also, suppose the second fashion product includes a logo. Also, suppose the composite information is a composite image including a model wearing the second fashion product. In this case, the acquisition unit 131 acquires information about the logo extracted from the second image as second extracted information. Note that the extraction process for extracting information about the logo from the second image can be realized by conventional techniques such as image analysis.

[0145] Furthermore, the determination unit 133 may use information about the logo to determine whether the composite image meets predetermined conditions. For example, the determination unit 133 may determine, based on at least the second image and information about the logo, whether the model is properly wearing the fashion product included in the composite image. This allows the determination unit 133 to determine with greater accuracy whether the composite information meets predetermined conditions.

[0146] [5-9. Provisioning Process] Furthermore, if the acquisition unit 131 has acquired the first and second composite elements that were already used to generate the composite information, the providing unit 134 may provide the composite information that has been generated before, which is the composite information obtained by combining the first and second composite elements. In this way, the providing unit 134 can reduce the occurrence of different output results each time composite information is generated. For example, suppose a user generates composite information multiple times using the first and second composite elements. In this case, the providing unit 134 may provide the same composite information each time.

[0147] Furthermore, the providing unit 134 may also provide the terminal device 10 with information regarding a download link that is valid for a certain period of time, as a way of providing the composite information. For example, when the user clicks the download link, the terminal device 10 transitions to a screen where the generated composite information can be downloaded. This allows the user to avoid, for example, permanently retaining images that include themselves.

[0148] [5-10. Reasons why the composite image does not meet the specified conditions] Furthermore, the reasons why the synthesized image does not meet the specified conditions may be generated by a large-scale language model. Here, a large-scale language model is a generative AI that generates character information. To give a more specific example, a large-scale language model is a model that has been trained to estimate and output the next token from an input sequence of tokens. For example, large-scale language models include transformer-based models and RNN (Recurrent Neural Network)-based models.

[0149] For example, a large-scale language model is a model trained to output a response corresponding to an input question, and is a language model that performs natural language processing such as GPT (Generative Pre-trained Transformer) or Transformer. The large-scale language model may reside within the information processing device 100 or on an external server. Furthermore, it is desirable for the large-scale language model to be trained so that the input information is not used as a new response, thereby concealing personal information and other sensitive data.

[0150] For example, suppose the external server has a large-scale language model. In this case, the determination unit 133 outputs information about areas that do not meet the predetermined conditions as information about whether the composite image meets the predetermined conditions. Subsequently, the determination unit 133 transmits information about areas that do not meet the predetermined conditions to the external server. In this case, the external server generates a reason using the large-scale language model. The determination unit 133 may then receive information about the reason from the external server and output such information about the reason.

[0151] [5-11. Prescribed conditions] The specified conditions may be changed depending on the user, product, or service content. For example, the specified conditions may be changed for each element such as color, pattern, design, silhouette, wrinkles, arrangement, or overall appearance, based on various trends of the user or shop (brand, product, etc.) or the content of the service. Furthermore, the specified conditions may be changed from one perspective to another. In addition, the specified conditions may be set from one perspective to another.

[0152] For example, the system may identify what aspects of fashion products the user considers important. For instance, the information processing device 100 may identify what aspects of fashion products the user considers important based on information regarding the user's returns of fashion products. For example, suppose the frequency with which the user returns products due to color differences exceeds a predetermined threshold. In this case, the information processing device 100 may identify that the user considers color important among fashion products based on information regarding the returns of fashion products.

[0153] Furthermore, the information processing device 100 may identify what the user considers important among fashion products based on the history of search queries entered by the user. For example, suppose the search query history includes search queries that search for details of colors such as jet black. In this case, the information processing device 100 may identify that the user considers color important among fashion products based on the search query history. In this way, the information processing device 100 identifies the keywords that the user considers important. This allows the information processing device 100 to modify predetermined conditions to suit the user, for example, based on the keywords that the user considers important. For example, by modifying predetermined conditions, the information processing device 100 can reduce the generation of composite images that alter color differences for users who do not consider color important.

[0154] [5-12. Provision of Synthesis Information] Furthermore, the invention is not limited to the above embodiment, and the providing unit 134 may provide the composite information if the determination unit 133 determines that the composite information satisfies predetermined conditions. In this case, the recipient of the composite information is a user who wants to obtain the composite information generated from the first composite element and the second composite element.

[0155] To give a more specific example, suppose the first composite element is a first image including a model wearing the first fashion product. Also, suppose the second composite element is a second image including the second fashion product. Also, suppose the composite information is a composite image including a model wearing the second fashion product. Also, suppose the re-composite information is a re-composite image obtained by re-composite the first image and the second image, and is a re-composite image including a model wearing the second fashion product. In this case, the acquisition unit 131 acquires the first image, the second image, and the composite image.

[0156] Next, the determination unit 133 determines, based on at least the second image, whether the composite image satisfies predetermined conditions. For example, the determination unit 133 determines that the composite image satisfies predetermined conditions if the model is properly wearing the second fashion product included in the composite image.

[0157] Furthermore, if the determination unit 133 determines that the composite image satisfies predetermined conditions, the providing unit 134 may provide the composite image to the recipient. In this way, the providing unit 134 can provide a suitable composite image that satisfies predetermined conditions.

[0158] Next, we will explain the case where it is determined that the composite image does not meet the predetermined conditions. In this case, if the determination unit 133 determines that the composite image does not meet the predetermined conditions, the acquisition unit 131 acquires a recomposite image that has been recompositely combined based on at least the second image, the composite image, and the determination result determined by the determination unit 133.

[0159] Here, the determination result includes information on whether the composite image satisfies predetermined conditions. For example, the determination result includes information on the score, the region of the composite image that does not satisfy the predetermined conditions, and at least one reason why the composite image does not satisfy the predetermined conditions. In this case, the generation unit 132 generates a recomposed image by inputting at least the second image, the composite image, and the determination result determined by the determination unit 133 to the first learning model 122. The generation unit 132 then stores the recomposed image in the image set storage unit 121. At this time, the acquisition unit 131 acquires the recomposed image from the image set storage unit 121.

[0160] Next, the determination unit 133 determines, based on at least the second image, whether the re-composite image satisfies predetermined conditions. For example, the determination unit 133 determines that the re-composite image satisfies predetermined conditions if the model is properly wearing the second fashion product included in the re-composite image.

[0161] Furthermore, the providing unit 134 may provide the re-composite image if the determination unit 133 determines that the re-composite image satisfies predetermined conditions. This allows the providing unit 134 to provide a re-composite image suitable for the recipient. In addition, the providing unit 134 can provide the recipient with a re-composite image with improved accuracy.

[0162] [5-13. Other Annotations] Furthermore, the information processing device 100 may provide information regarding the annotation tools described later. Here, for example, the first image IM11, the second image IM12, the third image IM13, and the composite image CI11 are assumed to be as follows.

[0163] For example, the first image IM11 includes the first model wearing the first fashion product. The second image IM12 includes the second fashion product. The third image IM13 includes the second model wearing the second fashion product. The composite image CI11 includes the first model wearing the second fashion product.

[0164] In such cases, in the above modified example, the annotator was made to evaluate whether the synthesized information (synthesized image CI11) met predetermined conditions (whether it was appropriate), but the large-scale language model and the annotator could also be made to evaluate whether the predetermined conditions were met.

[0165] For example, the information processing device 100 may input at least one of the following into a large-scale language model (second learning model 123, or a third learning model different from the first learning model 122 or second learning model 123): a first composite element (first image IM11), a second composite element (second image IM12), and reference information (third image IM13), along with the composite information, and have the large-scale language model generate regions that do not satisfy predetermined conditions or reasons why the predetermined conditions are not satisfied. The information processing device 100 may then present the regions that do not satisfy predetermined conditions or reasons why the predetermined conditions are not satisfied to an annotator, and have the annotator evaluate whether the regions that do not satisfy predetermined conditions or reasons why the predetermined conditions are not satisfied are appropriate.

[0166] The information processing device 100 may input information extracted for each evaluation perspective into the large-scale language model, or it may input all information into the large-scale language model regardless of the evaluation perspective.

[0167] Specifically, when evaluating from the first perspective of whether the appearance of the second fashion product included in the second composite element is maintained, the information processing device 100 may input the second composite element and the composite information into at least a large-scale language model and have the large-scale language model generate areas that do not satisfy predetermined conditions in the first perspective and the reasons why the predetermined conditions are not satisfied. The information processing device 100 may then present the areas that do not satisfy predetermined conditions in the first perspective and the reasons why the predetermined conditions are not satisfied, which were generated by the large-scale language model, to an annotator and have the annotator evaluate whether the areas that do not satisfy predetermined conditions and the reasons why the predetermined conditions are not satisfied are appropriate.

[0168] Furthermore, for example, the first composite element may further include a third fashion product worn by the first model in combination with the first fashion product. In this case, if the evaluation is to be performed from a second perspective, namely whether the appearance of the third fashion product combined with the first fashion product included in the first composite element is maintained, the information processing device 100 may input the first composite element and the composite information into at least a large-scale language model and have the large-scale language model generate areas that do not satisfy predetermined conditions in the second perspective and the reasons why the predetermined conditions are not satisfied. The information processing device 100 may then present the areas that do not satisfy predetermined conditions in the second perspective and the reasons why the predetermined conditions are not satisfied, generated by the large-scale language model, to an annotator and have the annotator evaluate whether the areas that do not satisfy predetermined conditions and the reasons why the predetermined conditions are not satisfied are appropriate.

[0169] Furthermore, if the evaluation is to be performed from a third perspective, namely whether the appearance of the first model included in the first composite element is maintained, the information processing device 100 may input the first composite element and the composite information into at least a large-scale language model and have the large-scale language model generate areas that do not satisfy predetermined conditions in the third perspective and the reasons why the predetermined conditions are not satisfied. The information processing device 100 may then present the areas that do not satisfy predetermined conditions in the third perspective and the reasons why the predetermined conditions are not satisfied, generated by the large-scale language model, to an annotator and have the annotator evaluate whether the areas that do not satisfy predetermined conditions and the reasons why the predetermined conditions are not satisfied are appropriate.

[0170] Furthermore, if evaluation is to be performed from a fourth perspective, which is whether the background contained in the first composite element is maintained, the information processing device 100 may input the first composite element and the composite information into at least a large-scale language model and have the large-scale language model generate areas that do not satisfy predetermined conditions in the fourth perspective and the reasons why the predetermined conditions are not satisfied. The information processing device 100 may then present the areas that do not satisfy predetermined conditions in the fourth perspective and the reasons why the predetermined conditions are not satisfied, generated by the large-scale language model, to an annotator and have the annotator evaluate whether the areas that do not satisfy predetermined conditions and the reasons why the predetermined conditions are not satisfied are appropriate.

[0171] Furthermore, when evaluation is performed from the first or second perspective, the information processing device 100 may also input auxiliary information generated for each perspective to be evaluated into the large-scale language model. Specifically, when evaluation is performed from the first perspective, the information processing device 100 may input auxiliary information about the second fashion product (such as the size of the second fashion product) generated from the second composite element and reference information into the large-scale language model. The information processing device 100 may also input at least the second composite element and reference information into the large-scale language model (the second learning model 123, the third learning model, the first learning model 122, the second learning model 123, or a fourth learning model different from the third learning model) and have the large-scale language model generate auxiliary information about the second fashion product.

[0172] Furthermore, when evaluation is performed from a second perspective, the information processing device 100 may input auxiliary information about the third fashion product (such as how to wear the third fashion product) generated from the first and second composite elements into a large-scale language model. The information processing device 100 may also input at least the first and second composite elements into a large-scale language model (such as the second learning model 123, the third learning model, the fourth learning model, or a fifth learning model different from the first learning model 122, the second learning model 123, the third learning model, or the fourth learning model) and have the large-scale language model generate auxiliary information about the third fashion product.

[0173] Furthermore, the information processing device 100 may have a large-scale language model and an annotator evaluate whether predetermined conditions are met in terms of suitability. For example, if a first model wearing a dress, which is the first fashion product, is to also wear a T-shirt, which is the second fashion product, then it is also necessary to have a fourth fashion product, bottoms (pants, skirt, etc.), also be worn. Specifically, when evaluating in terms of suitability, the information processing device 100 may input the synthesized information into a large-scale language model and have the large-scale language model generate areas that do not meet predetermined conditions in terms of suitability and the reasons why the predetermined conditions are not met. The information processing device 100 may then present the areas that do not meet predetermined conditions in terms of suitability and the reasons why the predetermined conditions are not met, which were generated by the large-scale language model, to the annotator and have the annotator evaluate whether the areas that do not meet predetermined conditions and the reasons why the predetermined conditions are not met are appropriate. In other words, the large-scale language model has learned coordinated images composed of multiple fashion products that have been evaluated as suitability in the past.

[0174] Furthermore, when evaluating from the perspective of suitability, the information processing device 100 may input auxiliary information related to suitability evaluation (such as whether the second fashion product and the fourth fashion product match, or whether the second fashion product and the fourth fashion product suit the first model) generated from the synthesized information into a large-scale language model.

[0175] Furthermore, auxiliary information regarding suitability evaluation may be generated using a learning model that, when a coordinated image is input, outputs evaluation results on whether the fashion products included in the coordinated image match each other, or evaluation results on whether one or more fashion products included in the coordinated image suit the model; a learning model that, when each image containing each fashion product that makes up the coordinated outfit is input, outputs evaluation results on whether the fashion products match each other; or a learning model that, when one or more images containing one or more fashion products that make up the coordinated outfit and a model image containing the model are input, outputs evaluation results on whether one or more fashion products suit the model.

[0176] Furthermore, the information processing device 100 may cause a large-scale language model to generate regions that satisfy predetermined conditions and reasons for satisfying those conditions, and have an annotator evaluate whether the regions that satisfy the predetermined conditions and reasons for satisfying those conditions are appropriate. In addition, the information processing device 100 may train the large-scale language model based on the evaluation results from the annotator.

[0177] [5-14. Program] Furthermore, the information processing device 100 according to the above-described embodiment is implemented by a computer 1000 having the configuration shown in Figure 12. Figure 12 is a hardware configuration diagram showing an example of a computer 1000 that implements the functions of the information processing device 100. The computer 1000 is connected to an output device 1010 and an input device 1020, and has a configuration in which an arithmetic unit 1030, a cache 1040, a memory 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 are connected by a bus 1090.

[0178] The arithmetic unit 1030 operates based on programs stored in the cache 1040 and memory 1050, as well as programs read from the input device 1020, and executes various processes. The cache 1040 is a memory device, such as RAM, that temporarily stores data used by the arithmetic unit 1030 for various calculations. The memory 1050 is a storage device where data used by the arithmetic unit 1030 for various calculations and various databases are registered, and is implemented using ROM (Read Only Memory), HDD (Hard Disk Drive), flash memory, etc.

[0179] Output IF1060 is an interface for transmitting information to be output to output devices 1010, such as monitors and printers, and is implemented using connectors of standards such as USB (Universal Serial Bus), DVI (Digital Visual Interface), and HDMI (High Definition Multimedia Interface). Input IF1070 is an interface for receiving information from input devices 1020, such as mice, keyboards, and scanners, and is implemented using, for example, USB.

[0180] The input device 1020 may also be a device that reads information from, for example, an optical recording medium such as a CD (Compact Disc), DVD (Digital Versatile Disc), or PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), tape media, magnetic recording media, or semiconductor memory. Furthermore, the input device 1020 may be an external storage medium such as a USB memory stick.

[0181] Network IF1080 receives data from other devices via network N and sends it to the arithmetic unit 1030, and also transmits data generated by the arithmetic unit 1030 to other devices via network N.

[0182] The arithmetic unit 1030 controls the output device 1010 and the input device 1020 via the output IF 1060 and the input IF 1070. For example, the arithmetic unit 1030 loads a program from the input device 1020 or memory 1050 onto the cache 1040 and executes the loaded program.

[0183] For example, when computer 1000 functions as an information processing device 100, the arithmetic unit 1030 of computer 1000 realizes the functions of the control unit 130 by executing a program loaded on the cache 1040.

[0184] [5-15. Others] Furthermore, among the processes described in the above embodiments and modifications, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically by known methods. In addition, the processing procedures, specific names, and information including various data and parameters shown in the above document and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each figure is not limited to the information shown.

[0185] Furthermore, the components of each illustrated device are functionally conceptual and do not necessarily need to be physically configured as shown. In other words, the specific forms of distribution and integration of each device are not limited to those shown, and all or part of them can be functionally or physically distributed and integrated in any unit according to various loads and usage conditions.

[0186] Furthermore, the embodiments and modifications described above can be combined as appropriate, provided that the processing content is not inconsistent.

[0187] Furthermore, the terms "section, module, unit" mentioned above can be replaced with "means" or "circuit," etc. For example, the acquisition unit can be replaced with acquisition means or acquisition circuit.

[0188] [6. Effects] As described above, the information processing device 100 according to the embodiment includes an acquisition unit 131, a determination unit 133, and a provision unit 134. The acquisition unit 131 acquires a first composite element, a second composite element, and composite information synthesized based on the first and second composite elements. The determination unit 133 determines, based on at least the second composite element, whether the composite information satisfies predetermined conditions. The provision unit 134 provides the composite information if the determination unit 133 determines that the composite information satisfies predetermined conditions.

[0189] The information processing device 100 according to this embodiment can provide suitable composite information that satisfies predetermined conditions.

[0190] Furthermore, in the information processing device 100 according to the embodiment, if the determination unit 133 determines that the synthesized information does not satisfy predetermined conditions, the acquisition unit 131 acquires re-synthesized information that has been re-synthesized based on at least a second synthesized element, the synthesized information, and the determination result determined by the determination unit 133. The determination unit 133 determines whether or not the re-synthesized information satisfies predetermined conditions based on at least the second synthesized element, and the provision unit 134 provides the re-synthesized information if the determination unit 133 determines that the re-synthesized information satisfies predetermined conditions.

[0191] As a result, the information processing device 100 according to the embodiment can provide suitable composite information that satisfies predetermined conditions.

[0192] Furthermore, in the information processing device 100 according to the embodiment, the determination unit 133 determines, based on at least the second synthesis element, whether or not the synthesis information satisfies a predetermined synthesis condition, which is the condition under which the first synthesis element and the second synthesis element are synthesized.

[0193] This allows the information processing device 100 according to the embodiment to determine whether the synthesized information has been properly synthesized.

[0194] Furthermore, in the information processing device 100 according to the embodiment, the determination unit 133 determines whether the composite information satisfies predetermined conditions, based on the relationship between the first composite element, the second composite element, the composite information, and related information of the first composite element, the second composite element, and the composite information, using a second learning model 123 that has been trained to output information on whether the composite information satisfies predetermined conditions when the first composite element, the second composite element, and the composite information are input.

[0195] This allows the information processing device 100 according to the embodiment to determine whether the synthesized information has been properly synthesized.

[0196] Furthermore, in the information processing device 100 according to the embodiment, the determination unit 133 determines whether the synthesized information satisfies predetermined conditions based on a predetermined evaluation policy.

[0197] This allows the information processing device 100 according to the embodiment to determine whether the synthesized information has been properly synthesized.

[0198] Furthermore, in the information processing device 100 according to the embodiment, the providing unit 134 further provides information corresponding to the determination result determined by the determination unit 133, which is information related to the synthesized information.

[0199] As a result, the information processing device 100 according to the embodiment can provide useful information for the composite image.

[0200] Furthermore, in the information processing device 100 according to the embodiment, the providing unit 134 provides information relating to a score corresponding to the determination result, which indicates the degree to which the composite information satisfies predetermined conditions.

[0201] As a result, the information processing device 100 according to the embodiment can provide useful information for the synthesized information.

[0202] Furthermore, in the information processing device 100 according to the embodiment, the providing unit 134 provides information about areas of the composite information that do not satisfy predetermined conditions, based on the determination result.

[0203] As a result, the information processing device 100 according to the embodiment can provide useful information for the synthesized information.

[0204] Furthermore, in the information processing device 100 according to the embodiment, the providing unit 134 provides information on the reason why the synthesized information does not meet predetermined conditions, based on the determination result.

[0205] As a result, the information processing device 100 according to the embodiment can provide useful information for the synthesized information.

[0206] Furthermore, in the information processing device 100 according to the embodiment, the acquisition unit 131 further acquires the first extracted information extracted from the first composite element, and the determination unit 133 uses the first extracted information to determine whether or not the composite information satisfies predetermined conditions.

[0207] This allows the information processing device 100 according to the embodiment to determine whether the synthesized information has been properly synthesized.

[0208] Furthermore, in the information processing device 100 according to the embodiment, the acquisition unit 131 further acquires the second extracted information extracted from the second composite element, and the determination unit 133 uses the second extracted information to determine whether the composite information satisfies predetermined conditions.

[0209] This allows the information processing device 100 according to the embodiment to determine whether the synthesized information has been properly synthesized.

[0210] Furthermore, in the information processing device 100 according to the embodiment, the acquisition unit 131 acquires a first composite element as a first composite element, in which the region for compositing the second composite element is masked from the first composite element.

[0211] As a result, the information processing device 100 according to the embodiment can acquire suitable data for determining whether the synthesized information has been properly synthesized.

[0212] Furthermore, in the information processing device 100 according to the embodiment, the acquisition unit 131 acquires a first image including a person wearing the first fashion product as a first composite element, a second image including the second fashion product as a second composite element, and a composite image formed based on the first image and the second image, which includes a person wearing the second fashion product, and the determination unit 133 determines whether the composite image satisfies predetermined conditions.

[0213] This allows the information processing device 100 according to the embodiment to determine whether the synthesized information has been properly synthesized.

[0214] Furthermore, the information processing device 100 according to the embodiment further includes a generation unit 132 that generates a second learning model 123 which is trained to output information on whether or not the composite information satisfies predetermined conditions when the first composite element, the second composite element, and the composite information are input, based on the relationship between the first composite element, the second composite element, the composite information, and related information of the first composite element, the second composite element, and the composite information.

[0215] As a result, the information processing device 100 according to the embodiment can generate a suitable second learning model 123 for determining whether the synthesized information has been properly synthesized.

[0216] Furthermore, the information processing device 100 according to the embodiment further includes a learning unit 136 that teaches the first learning model 122, which has been trained to output synthesized information when a first synthesized element and a second synthesized element are input, the relationship between the first synthesized element, the second synthesized element, the synthesized information, and the determination result determined by the determination unit 133.

[0217] As a result, the information processing device 100 according to the embodiment can improve the output accuracy of the first learning model 122.

[0218] Although some embodiments of the present invention have been described in detail above with reference to the drawings, these are illustrative examples, and the present invention can be implemented in various other forms with modifications and improvements based on the knowledge of those skilled in the art, starting with the embodiments described in the disclosure section of the invention. [Explanation of Symbols]

[0219] N Network 1. Information Processing System 10 Terminal devices 100 Information Processing Devices 110 Communications Department 120 Storage section 121 Image set storage unit 122 First Learning Model 123 Second Learning Model 124 Provided information storage unit 130 Control Unit 131 Acquisition Department 132 Generation part 133 Judgment section 134 Provision Department 135 Reception Department 136 Learning Department

Claims

1. An acquisition unit that acquires a first synthesis element, a second synthesis element, and synthesis information synthesized based on the first and second synthesis elements, A determination unit that determines whether the composite information satisfies predetermined conditions based on at least the second composite element, If the determination unit determines that the composite information satisfies the predetermined conditions, the providing unit that provides the composite information An information processing device characterized by comprising:

2. The acquisition unit is, If the determination unit determines that the synthesis information does not satisfy the predetermined conditions, it obtains the re-synthesized information, which is re-synthesized based on at least the second synthesis element, the synthesis information, and the determination result determined by the determination unit. The determination unit, Based on at least the second synthesis element, it is determined whether the resynthesis information satisfies the predetermined conditions. The aforementioned supply unit is, If the determination unit determines that the recombination information satisfies the predetermined conditions, it provides the recombination information. The information processing apparatus according to feature 1.

3. The determination unit, Based at least the second composite element, it is determined whether the composite information satisfies the predetermined conditions, which are the conditions under which the first composite element and the second composite element are combined. The information processing apparatus according to feature 1.

4. The determination unit, Using a learning model trained to output information on whether the composite information satisfies predetermined conditions when the first composite element, the second composite element, and the composite information are input, based on the relationship between the first composite element, the second composite element, the composite information, and related information of the composite information, the determination of whether the composite information satisfies predetermined conditions is made using a learning model that has been trained to output information on whether the composite information satisfies predetermined conditions. The information processing apparatus according to feature 1.

5. The determination unit, Based on a predetermined evaluation policy, it is determined whether the synthesized information satisfies predetermined conditions. The information processing apparatus according to feature 1.

6. The aforementioned supply unit is, Information corresponding to the determination result determined by the determination unit, which further provides information on the synthesized information. The information processing apparatus according to feature 1.

7. The aforementioned supply unit is, The present invention provides information relating to a score corresponding to the judgment result, the score indicating the degree to which the composite information satisfies the predetermined conditions. The information processing apparatus according to feature 6.

8. The aforementioned supply unit is, Based on the above determination result, information is provided regarding the region of the composite information that does not satisfy the predetermined conditions. The information processing apparatus according to feature 6.

9. The aforementioned supply unit is, Based on the above determination result, information is provided regarding the reason why the composite information does not meet the predetermined conditions. The information processing apparatus according to feature 6.

10. The acquisition unit is, Further obtaining the first extracted information extracted from the first composite element, The determination unit, The first extracted information is used to determine whether the composite information satisfies predetermined conditions. The information processing apparatus according to feature 1.

11. The acquisition unit is, Further obtaining the second extracted information extracted from the second composite element, The determination unit, The second extracted information is used to determine whether the composite information satisfies predetermined conditions. The information processing apparatus according to feature 1.

12. The acquisition unit is, As the first composite element, the first composite element is obtained in which the region of the first composite element that combines the second composite element is masked. The information processing apparatus according to feature 1.

13. The acquisition unit is, The first composite element is a first image including a person wearing the first fashion product, the second composite element is a second image including the second fashion product, and the composite image obtained by combining the first image and the second image is a composite image including the person wearing the second fashion product. The determination unit, Determining whether the composite image satisfies predetermined conditions The information processing apparatus according to feature 1.

14. The system further comprises a generation unit that generates a learning model that, based on the relationship between the first synthesis element, the second synthesis element, the synthesis information, and related information of the first synthesis element, the second synthesis element, and the synthesis information, outputs information regarding whether the synthesis information satisfies predetermined conditions when the first synthesis element, the second synthesis element, and the synthesis information are input. The information processing apparatus according to feature 1.

15. The system further includes a learning unit that learns the relationship between the first composite element, the second composite element, the composite information, and the determination result determined by the determination unit, to a learning model that has been trained to output the composite information when the first composite element and the second composite element are input. The information processing apparatus according to feature 1.

16. A method of information processing performed by a computer, An acquisition step of acquiring a first composite element, a second composite element, and composite information synthesized based on the first and second composite elements, A determination step of determining whether the composite information satisfies predetermined conditions based on at least the second composite element, If the determination step determines that the synthesis information satisfies the predetermined conditions, the provision step provides the synthesis information. An information processing method characterized by including

17. An acquisition procedure for acquiring a first composite element, a second composite element, and composite information synthesized based on the first and second composite elements, A determination procedure for determining whether the composite information satisfies predetermined conditions based on at least the second composite element, If the determination procedure determines that the composite information satisfies the predetermined conditions, then the procedure for providing the composite information is performed. An information processing program characterized by causing a computer to execute it.

Images