JP2019153223A - Device, method, and program for generating images - Google Patents

Abstract

To allow for easily providing cooking images which present a room for a creative touch that may help create variation-rich dishes.SOLUTION: An image generation device 10 acquires information on cooking ingredients entered by a user and random number data generated at random. The image generation device 10 then uses the acquired information on the cooking ingredients and random number data as input to generate a cooking image using a learned model for generating cooking images.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an image generation apparatus, an image generation method, and an image generation program.

  Traditionally, in home cooking, while using the ingredients you already have, you want to ensure appropriate diversity in the target dishes in order to improve the quality of life, or you want to create new dishes by ingenuity Situations often occur. In such a situation, there is a technique that suggests the direction of cooking that can be made with ingredients at hand to the user.

  As such a technique, for example, a search technique (for example, Cookpad (registered trademark)) on a database storing foods corresponding to a cooking image or a general-purpose web image search technique (for example, Google (registered trademark) image search). and so on. In such a search technique, when a search query such as a dish name or an ingredient name is input, a search result is output according to popularity and relevance.

  There is also an automatic recipe generation technology (for example, Chef Watson (registered trademark)) based on classical knowledge processing or natural language processing technology. In such a technique, an existing recipe is modified to output a new recipe, and an existing dish image corresponding to the existing recipe is output.

JP 2004-192050 A

  However, the conventional method has a problem in that it is not possible to easily provide a cooking image that has room for creative ideas that can help a variety of dishes. For example, in the conventional search technique described above, the output search results are ordered by popularity and relevance, so that only similar dishes tend to be displayed at the top, and in order to obtain appropriate and diverse search results However, there is a problem that the user side is required to create a search query. In addition, because of the search technique, there is a problem that the user cannot be presented with anything other than the dishes existing on the database or on the web.

  In addition, the conventional recipe automatic generation technique described above modifies an existing recipe, so that a recipe that is fundamentally different from the original recipe is not generated, and the recipe is specific and detailed. Therefore, there is a problem that there is little room for the user to make ingenuity. In addition, because the dish image output with the modified recipe is an existing dish image, the appearance of the dish created by the new recipe is not presented, and it is difficult to guess the appearance of the dish until it is actually cooked .

  In order to solve the above-described problems and achieve the object, an image generation apparatus according to the present invention includes an acquisition unit that acquires information about foods input by a user and randomly generated random number data, and the acquisition unit And a learned generation unit that generates the dish image using the learned model for generating a dish image by using the information on the ingredients acquired by the above and the random number data as inputs.

  Further, the image generation method of the present invention is an image generation method executed by an image generation apparatus, and obtains information related to ingredients input by a user and randomly generated random number data, And a learned generation step of generating the dish image using a learned model for generating a dish image by using the information on the ingredients acquired in the acquisition step and the random number data as inputs. To do.

  In addition, the image generation program of the present invention includes an acquisition step of acquiring information related to a food material input by a user and randomly generated random number data, and information related to the food material acquired by the acquisition step and the random number data. And a learned generation step for generating the cooking image using a learned model for generating a cooking image.

  According to the present invention, there is an effect that it is possible to easily provide a cooking image that has a room for inventiveness that helps a variety of dishes.

FIG. 1 is a block diagram illustrating a configuration example of an image generation apparatus according to the first embodiment. FIG. 2 is a diagram illustrating an example of data stored in the dish data storage unit. FIG. 3 is a diagram for explaining learning processing in the image generation apparatus according to the first embodiment. FIG. 4 is a diagram illustrating an example of a food input screen displayed on the user terminal. FIG. 5 is a diagram for explaining image generation processing in the image generation apparatus according to the first embodiment. FIG. 6 is a diagram illustrating an example of a cooking image output screen displayed on the user terminal. FIG. 7 is a flowchart illustrating an example of a learning process in the image generation apparatus according to the first embodiment. FIG. 8 is a flowchart illustrating an example of the flow of image generation processing in the image generation apparatus according to the first embodiment. FIG. 9 is a diagram illustrating a computer that executes an image generation program.

  Hereinafter, embodiments of an image generation apparatus, an image generation method, and an image generation program according to the present application will be described in detail with reference to the drawings. Note that the embodiment does not limit the image generation apparatus, the image generation method, and the image generation program according to the present application.

[First Embodiment]
In the following embodiments, the configuration of the image generation device 10 according to the first embodiment and the flow of processing of the image generation device 10 will be described in order, and finally the effects of the first embodiment will be described.

[Configuration of Image Generation Device]
FIG. 1 is a block diagram illustrating a configuration example of an image generation apparatus according to the first embodiment. The configuration of the image generation apparatus 10 will be described with reference to FIG. As shown in FIG. 1, the image generation apparatus 10 is connected to a user terminal 20 via a network 30.

  Here, the user terminal 20 is an information processing apparatus such as a desktop PC, a tablet PC, a notebook PC, a mobile phone, a smartphone, or a PDA (Personal Digital Assistant).

  As illustrated in FIG. 1, the image generation apparatus 10 includes a communication processing unit 11, a control unit 12, and a storage unit 13. Hereinafter, processing of each unit included in the image generation apparatus 10 will be described.

  The communication processing unit 11 controls communication related to various types of information. For example, the communication processing unit 11 receives an ingredient name and a cooking image generation request from the user terminal 20. In addition, the communication processing unit 11 transmits the generated cooking image to the user terminal 20.

  The storage unit 13 stores data and programs necessary for various processes performed by the control unit 12, and particularly includes a dish data storage unit 13 a that is closely related to the present invention. For example, the storage unit 13 is a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The data stored in the dish data storage unit 13a is data stored in advance and can be updated as appropriate.

  The dish data storage unit 13a stores a dish image, which is an image of a genuine dish, and information related to the ingredients used for the genuine dish in association with each other. For example, as illustrated in FIG. 2, the dish data storage unit 13 a includes a “cooking image” that is an image of a real dish, a “food name” that indicates the name of the ingredient used for cooking the dish image, and an ingredient. The “food vector” obtained by vectorizing the name is stored in association with each other. FIG. 2 is a diagram illustrating an example of data stored in the dish data storage unit.

  2, the dish data storage unit 13a associates the dish image “image A”, the ingredient name “egg, chicken, onion...” And the ingredient vector “vector A”. Remember. The information illustrated in FIG. 2 is an example, and the present invention is not limited to this. In the example of FIG. 2, the dish image and the ingredient vector are simply described as image A, vector A, and the like.

  The control unit 12 has an internal memory for storing a program that defines various processing procedures and necessary data, and performs various processes using them, and particularly as closely related to the present invention, A generation unit 12a, an identification unit 12b, a learning unit 12c, an acquisition unit 12d, and a learned generation unit 12e are included. Here, the control unit 12 is an electronic circuit such as a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or a GPU (Graphical Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or the like. Integrated circuit.

  The processing executed by each functional unit of the control unit 12 is roughly divided into a learning phase in which machine learning of a model is performed and an image generation phase in which a cooking image is generated using a learned model. The generation unit 12a, the identification unit 12b, and the learning unit 12c in the control unit 12 are functional units that perform processing in the learning phase, and the acquisition unit 12d and the learned generation unit 12e are functional units that perform processing in the image generation phase. . Note that the image generation apparatus 10 according to the first embodiment will be described as an apparatus that performs both the learning process in the learning phase and the image generation process in the image generation phase, but is not limited to this, and the image generation phase Only the image generation process in the above may be performed. In this case, it is assumed that the image generation apparatus 10 has previously set a learned model that has been subjected to machine learning processing in advance.

  In the learning phase, for example, a GAN (Generative Adversarial Network), which is a kind of a neural network, is used to combine two neural networks, a generator and a discriminator, for a given data set. Do learning. For example, as a learning process, when the target data is an image, the discriminator generates whether the input image belongs to the original data set so that the generator generates some random image. Constructed to identify what is. The learning process in the learning phase and the image generation process in the image generation phase will be described in detail later with reference to the drawings. Below, each function part is demonstrated.

  The generation unit 12a receives information about the ingredients stored in the dish data storage unit 13a and a random number as inputs, and uses the first model (hereinafter referred to as “generator”) to generate a dish image, and uses the dish image. Is generated.

  A specific process will be described. The generation unit 12a performs random sampling from the dish data storage unit 13a to acquire a set of food names. And the production | generation part 12a converts the acquired foodstuff name into a vector. For example, the generation unit 12a performs word-embedding on the acquired food name to convert it into a vector φ, and further expresses the vector φ as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)). Dimension compression is performed. Then, the generation unit 12a combines the vector c ′ and the random number vector z, and generates a dish image using the combined vector as an input.

  The identification unit 12b uses, as training data, information about the cooking image and ingredients stored in the cooking data storage unit 13a as a second model for identifying the authenticity of the input image (hereinafter referred to as “discriminator”). Is used to identify the authenticity of the dish image generated by the generation unit 12a.

  Explaining the specific processing, the identification unit 12b includes a cooking image generated by the generator, a real cooking image corresponding to the food name obtained by the sampling unit 12a performing random sampling from the cooking data storage unit 13a, and The food vector c ′ is input to the discriminator, and a convolutional neural network is used to discriminate whether the dish image generated by the generator is genuine. Note that any existing method may be used as a method for identifying how it looks authentic. For example, the distance of the probability distribution between the dish image generated by the generator and the genuine dish image is determined. You may make it calculate and define authenticity according to the calculated distance.

  The learning unit 12c optimizes the generator so that a genuine dish image can be generated, and optimizes the classifier so that the dish image identification accuracy can be improved. For example, as described above, each time the learning unit 12c generates a cooking image by the generator, and each time the authenticity of the cooking image is identified by the classifier, the learning unit 12c sets the parameters of the generator so that a genuine cooking image can be generated. The parameters of the discriminator are optimized so that the cooking image discrimination accuracy can be improved. Note that the learning unit 12c may use any method for optimizing the parameters, and may apply any of the existing optimization methods in machine learning.

  Here, a series of learning processes in the image generating apparatus 10 according to the first embodiment will be described with reference to FIG. FIG. 3 is a diagram for explaining learning processing in the image generation apparatus according to the first embodiment. As shown in FIG. 3, the image generating apparatus 10 applies a combination of two neural networks, ie, a generator and a discriminator, so that the generator generates a random dish image, and the discriminator inputs the dish image that is input Is configured to identify whether it belongs to a real cooking image or a cooking image generated by the generator.

  As illustrated in FIG. 3, the image generation apparatus 10 performs “word-embedding” on the “food material name” acquired from the dish data storage unit 13 a and converts it into a vector φ. Then, the image generation apparatus 10 performs dimensional compression by expressing the vector φ as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)), and combines the vector c ′ and the random number vector z. (Refer to FIG. 3A).

  Then, the image generation apparatus 10 inputs the combined vector to the generator, and generates a dish image by upsampling using a convolutional neural network (see FIG. 3B).

  Subsequently, the image generation apparatus 10 generates the cooking image generated by the generator, and the genuine cooking image and the food vector c ′ corresponding to the food name acquired by the generation unit 12a performing random sampling from the food data storage unit 13a. Are input to the discriminator, and a convolutional neural network is used to discriminate whether the dish image generated by the generator is genuine (see FIG. 3C).

  The image generating apparatus 10 repeats the above processing (the processing of (A) to (C) in FIG. 3), and the discriminator improves the discrimination accuracy of the cooking image so that the generator can generate a more authentic cooking image. Optimize the parameters of each neural network as possible. In this way, in the image generation device 10, by training two neural networks simultaneously, if learning is successful, the generator of the learned model randomly generates a dish image that is difficult to distinguish from a real dish image. Will be output.

  That is, the image generation apparatus 10 uses a combination of actual dish images and the names of ingredients used in the dish as a data set used for learning. Then, when the generator generates an image at random and when the discriminator identifies the input image, the image generation device 10 uses the names of ingredients used for cooking each image as additional information. Improve the structure of GAN to use. As a result, the trained model generator randomly generates and outputs corresponding dish images given the names of the ingredients.

  Further, in the image generation apparatus 10, after “word-embedding” is performed on the “food name” to convert it into a vector φ, the vector φ is set as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)). expressing. Normally, the types of ingredients are enormous (hundreds of dimensions), and if the inputted ingredient name is converted into a normal one-hot vector or the like, the expression ability is lowered. For this reason, if the One-hot vector is input to the generator or discriminator as it is, many elements of the vector are collapsed to 0, and learning may not progress well.

  On the other hand, the image generating apparatus 10 performs word-embedding on the “food name” and converts it into a vector φ, and then the vector φ is a latent variable based on the normal distribution N (μ (φ), σ (φ)). By expressing as c ′, higher expression capability is converted into a lower dimensional vector. As a result, it is possible to prevent many of the elements from being crushed to 0 and to facilitate the learning. The latent variable c ′ is an expression based on the probability distribution and does not impair the diversity of images generated by the generator.

  Returning to the description of FIG. 1, the acquisition unit 12 d acquires information on the ingredients input by the user and randomly generated random data. For example, the acquisition unit 12d acquires the name of the ingredient input on the input screen displayed on the user terminal 20. Here, an example of the input screen of the foodstuff displayed on the user terminal 20 is demonstrated using the example of FIG. FIG. 4 is a diagram illustrating an example of a food input screen displayed on the user terminal.

  As illustrated in FIG. 4, the user terminal 20 displays a text box for inputting an ingredient name and a button for instructing generation of a cooking image. For example, as illustrated in FIG. 4, in the input screen displayed on the user terminal 20, enter “egg, chicken, onion ...” in the text box, and press the button displayed as “image generation”. Thus, the ingredient name “egg, chicken, onion...” And a cooking image generation request are transmitted from the user terminal 20 to the image generation apparatus 10. As an example of the situation at the time of input, for example, when a user prepares dinner or the like, the name of the ingredients at hand is input into a text box.

  The input screen illustrated in FIG. 4 is an example, and the present invention is not limited to this. For example, the example of FIG. 4 shows an example in which one text box is displayed and a plurality of food names are input in one text box, but a plurality of text boxes are displayed, one in each text box. An ingredient name may be input. Moreover, it may replace with a text box and may display the pull-down list for selecting a foodstuff name, and may enable it to select a foodstuff from a pull-down list.

  Moreover, the acquisition part 12d will convert a foodstuff name into a vector, if a foodstuff name is received from the user terminal 20. FIG. For example, the acquisition unit 12d performs word-embedding on the acquired ingredient name to convert it into a vector φ, and further expresses the vector φ as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)). Dimension compression is performed. Then, the acquisition unit 12d combines the food vector c ′ and the random number vector z.

  The learned generation unit 12e uses the learned model for generating a cooking image, using the information about the food and the random number data acquired by the acquisition unit 12d as inputs, and generates a cooking image. Specifically, the learned generation unit 12e generates a dish image using the learned model (generator) with the vector obtained by combining the food vector c ′ and the random number vector z by the acquisition unit 12d as an input. That is, the learned generation unit 12e generates a dish image using the generator optimized by the learning unit 12c as a learned model. Thereafter, the learned generation unit 12e outputs the generated dish image to the user terminal 20.

  Here, a series of flow of image generation processing in the image generation apparatus 10 according to the first embodiment will be described with reference to FIG. FIG. 5 is a diagram for explaining image generation processing in the image generation apparatus according to the first embodiment. As shown in FIG. 5, in the image generation device 10, two neural networks of a generator and a discriminator are applied in combination, and the discriminator inputs the dish image inputted so that the generator generates a random dish image. Is configured to identify whether it belongs to a real cooking image or a cooking image generated by the generator.

  As illustrated in FIG. 5, the image generation apparatus 10 performs “word-embedding” on the “food material name” acquired from the dish data storage unit 13 a and converts it into a vector φ. Then, the image generation apparatus 10 performs dimensional compression by expressing the vector φ as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)), and combines the vector c ′ and the random number vector z. To do.

  Then, the image generation apparatus 10 inputs the combined vector to a generator that is a learned model, and generates a dish image by upsampling using a convolutional neural network. Thereafter, the image generation apparatus 10 transmits the generated dish image to the user terminal 20.

  Here, the output screen of the dish image displayed on the user terminal 20 will be described using the example of FIG. FIG. 6 is a diagram illustrating an example of a cooking image output screen displayed on the user terminal. As illustrated in FIG. 6, in the user terminal 20, a dish image is displayed as a dish image output screen under a text box in which the name of the ingredient is input. The dish image displayed here is randomly generated by the learned model generator. For example, the user terminal 20 displays a dish image different from the existing dish image, which is randomly generated with the name of the ingredient at hand at the user.

  In this way, the image generation apparatus 10 randomly generates a cooking image by using the learned model generator. For example, even when the food at the user's hand is the same as before, there is a corresponding variation, and the existing cooking A dish image different from the image can be provided to the user. For this reason, it is possible to easily obtain various and novel suggestions about cooking that the user can make using the ingredients at hand in cooking at home, which can lead to improvement of the quality of life through its ingenuity. it can.

[Processing procedure of image generating apparatus]
Next, an example of a processing procedure performed by the image generation apparatus 10 according to the first embodiment will be described with reference to FIGS. 7 and 8. FIG. 7 is a flowchart illustrating an example of a learning process in the image generation apparatus according to the first embodiment. FIG. 8 is a flowchart illustrating an example of the flow of image generation processing in the image generation apparatus according to the first embodiment.

  First, an example of a learning process flow in the image generation apparatus 10 will be described with reference to FIG. As illustrated in FIG. 7, the generation unit 12a of the image generation device 10 performs random sampling from the dish data storage unit 13a to acquire the ingredient name (step S101). And the production | generation part 12a performs Word-embedding for the acquired foodstuff name, and converts it into vector (phi) (step S102).

  Subsequently, the generation unit 12a performs dimensional compression by expressing the converted vector φ as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)) (step S103). Then, the generation unit 12a combines the vector c ′ and the random number vector z (step S104), and generates a dish image from the combined vector using the convolutional neural network of the generator (step S105).

  And the identification part 12b produces | generates the cooking image produced | generated by the generator, and the real cooking image and ingredients vector c 'corresponding to the foodstuff name which the production | generation part 12a acquired by performing random sampling from the dish data storage part 13a. It inputs into a discriminator, and discriminate | determines whether the dish image produced | generated by the generator is genuine using the convolution neural network of a discriminator (step S106).

  Thereafter, the learning unit 12c optimizes the parameters of the generator so as to generate a genuine dish image by a predetermined method, and optimizes the parameters of the classifier so that the dish image identification accuracy can be improved (step S107). ). Note that the image generation apparatus 10 repeats the series of processing in steps S101 to S107 until a predetermined condition is satisfied. For example, the image generation apparatus 10 may set the number of repetitions in advance, or may be repeated until the accuracy of the generator and the discriminator satisfies a predetermined threshold.

  Next, an example of the flow of image generation processing in the image generation apparatus 10 will be described with reference to FIG. As illustrated in FIG. 8, when the acquisition unit 12d receives an input of a food name from the user terminal 20 (Yes in Step S201), the acquisition unit 12d converts the food name into a vector φ (Step S202). Subsequently, the acquisition unit 12d performs dimensional compression by expressing the converted vector φ as a latent variable c ′ based on the normal distribution N (μ (φ), σ (φ)) (step S203). Then, the acquisition unit 12d combines the food vector c ′ and the random number vector z (step S204).

  Then, the learned generation unit 12e generates a dish image by using the learned vector (generator) with the combined vector as an input (step S205). Thereafter, the learned generation unit 12e outputs the generated dish image to the user terminal 20 (step S206).

(Effects of the first embodiment)
The image generation apparatus 10 according to the first embodiment acquires information on food ingredients input by the user and randomly generated random number data, and inputs the information on the acquired food ingredients and random number data, and inputs a dish image. A cooking image is generated using the learned model for generation. For this reason, it is possible to easily provide a cooking image with room for ingenuity that helps a variety of dishes.

  In other words, the image generation device 10 can generate a randomly unknown and genuine dish image by the learned model generator, and there is a corresponding variation even when the ingredients at the user's hand are the same or similar as before. Yes, it is possible to provide a cooking image that allows the user to imagine the appearance of the cooking to some extent.

  In addition, since the image generation apparatus 10 does not present a recipe but provides a user with a cooking image that is different from the existing cooking image, there is room for a certain amount of ingenuity in cooking. Moreover, since the image generation apparatus 10 only needs to input the name of the foodstuff at hand, it can obtain a cooking image easily. As described above, the image generation apparatus 10 can easily obtain various and novel suggestions about the dishes that the user can make using the ingredients at hand in cooking at home, It is possible to improve the quality.

(System configuration etc.)
Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. Further, all or any part of each processing function performed in each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.

  In addition, among the processes described in this embodiment, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed All or a part of the above can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

(program)
In addition, it is possible to create a program in which processing executed by the image generation apparatus described in the above embodiment is described in a language that can be executed by a computer. For example, it is possible to create an image generation program in which processing executed by the image generation apparatus 10 according to the embodiment is described in a language that can be executed by a computer. In this case, when the computer executes the image generation program, it is possible to obtain the same effect as in the above embodiment. Further, the image generation program may be recorded on a computer-readable recording medium, and the image generation program recorded on the recording medium may be read by the computer and executed to execute the same processing as in the above embodiment. Good.

  FIG. 9 is a diagram illustrating a computer that executes an image generation program. As illustrated in FIG. 9, the computer 1000 includes, for example, a memory 1010, a CPU 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These units are connected by a bus 1080.

  The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012 as illustrated in FIG. The ROM 1011 stores a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1090 as illustrated in FIG. The disk drive interface 1040 is connected to the disk drive 1100 as illustrated in FIG. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example, as illustrated in FIG. The video adapter 1060 is connected to a display 1130, for example, as illustrated in FIG.

  Here, as illustrated in FIG. 9, the hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the above-described image generation program is stored in, for example, the hard disk drive 1090 as a program module in which a command to be executed by the computer 1000 is described.

  In addition, various data described in the above embodiment is stored as program data in, for example, the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads the program module 1093 and the program data 1094 stored in the memory 1010 and the hard disk drive 1090 to the RAM 1012 as necessary, and executes various processing procedures.

  Note that the program module 1093 and the program data 1094 related to the image generation program are not limited to being stored in the hard disk drive 1090, but are stored in, for example, a removable storage medium and read out by the CPU 1020 via the disk drive or the like. Also good. Alternatively, the program module 1093 and the program data 1094 related to the image generation program are stored in another computer connected via a network (LAN (Local Area Network), WAN (Wide Area Network), etc.), and the network interface 1070 is stored. Via the CPU 1020.

  The above embodiments and modifications thereof are included in the invention disclosed in the claims and equivalents thereof as well as included in the technology disclosed in the present application.

DESCRIPTION OF SYMBOLS 10 Image generation apparatus 11 Communication processing part 12 Control part 12a Generation part 12b Identification part 12c Learning part 12d Acquisition part 12e Learned generation part 13 Storage part 13a Cooking data storage part 20 User terminal 30 Network

Claims (5)

An acquisition unit that acquires information about ingredients input by the user and randomly generated random number data;
A learning generation unit that generates the cooking image using a learned model for generating a cooking image by using the information on the ingredients acquired by the acquisition unit and the random number data as inputs. An image generating device. A storage unit that stores a cooking image that is an image of a real dish and information related to the ingredients used in the real cooking in association with each other;
A generation unit that generates the dish image using a first model that generates information about the food stored in the storage unit and the random number data and generates a dish image;
Using the second model for identifying the authenticity of the input image as training data, the authenticity of the cooking image generated by the generating unit using the cooking image and the information on the ingredients stored in the storage unit as training data. An identification unit for identifying
A learning unit that optimizes the first model so as to generate a genuine dish image and optimizes the second model so that the identification accuracy of the dish image can be improved;
The image generation apparatus according to claim 1, wherein the learned generation unit generates the dish image using the first model optimized by the learning unit as the learned model.   The generation unit performs word-embedding on the food name as information on the food stored in the storage unit, converts the name into a vector, and performs dimensional compression by expressing the vector as a latent variable based on a normal distribution. 3. The image generation according to claim 2, wherein the dish image is generated by combining the dimension-compressed vector and the random number vector, and using the combined vector as an input, using the first model. apparatus. An image generation method executed by an image generation apparatus,
An acquisition step of acquiring information about ingredients input by the user and randomly generated random number data;
And a learned generation step of generating the dish image using the learned model for generating a dish image using the information on the ingredients acquired in the acquisition step and the random number data as inputs. An image generation method. An acquisition step of acquiring information about ingredients input by the user and randomly generated random number data;
Causing the computer to execute a learned generation step of generating the dish image using the learned model for generating a dish image by using the information on the ingredients acquired in the acquisition step and the random number data as inputs. An image generation program characterized by the above.

Images