JP7740446B2 - Information processing device, information processing method, and program - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and a program.

Various methods have been proposed to assist in creating layouts for magazines, books, newspapers, etc. For example, Patent Document 1 discloses an information processing device that calculates the degree of relevance between content such as text and images, which are layout materials, and determines a layout so that content with a high degree of relevance is placed close to each other.

JP 2009-169536 A

However, the invention of Patent Document 1 merely places highly related content close to each other, and there is a risk that the layout may not be appropriate when viewed as a whole.

One aspect is to provide an information processing device or the like that can generate an appropriate layout.

In one aspect, the information processing device includes an acquisition unit that acquires multiple pieces of content, a generation unit that generates a layout image in which the multiple pieces of content are arranged within a predetermined area, an evaluation unit that acquires an evaluation of the generated layout image using a classifier that has trained on the multiple layout images, and an output unit that outputs the evaluation results.

On the one hand, it can generate an appropriate layout.

FIG. 1 is a schematic diagram illustrating an example of the configuration of a layout generation system. FIG. 2 is a block diagram illustrating an example of the configuration of a server. FIG. 1 is an explanatory diagram showing an overview of the present embodiment. FIG. 10 is an explanatory diagram relating to layout learning processing. FIG. 10 is an explanatory diagram relating to a layout generation process. 10 is a flowchart illustrating an example of a processing procedure for layout learning processing. 10 is a flowchart illustrating an example of a processing procedure for a layout generation process. FIG. 10 is an explanatory diagram showing an overview of a second embodiment. 10 is a flowchart illustrating an example of a processing procedure for a layout generation process according to the second embodiment. FIG. 2 is a functional block diagram showing the operation of the server of the above-described embodiment. FIG. 1 is an explanatory diagram showing the main parts of a scoring model 141. FIG. 10 is an explanatory diagram showing an example of calculation of a total score. FIG. 10 is an explanatory diagram showing an overview of a fourth embodiment. 13 is a flowchart showing an example of a processing procedure for a layout generation process according to the fourth embodiment.

The present invention will be described in detail below with reference to the drawings showing embodiments thereof.
(Embodiment 1)
1 is a schematic diagram showing an example of the configuration of a layout generation system. In this embodiment, a layout generation system that automatically generates page layouts for documents such as magazines, books, newspapers, etc. will be described. The layout generation system includes an information processing device 1 and a terminal 2. Each device is communicatively connected via a network N such as the Internet.

The information processing device 1 is an information processing device capable of various information processing and information transmission and reception, such as a server device or a personal computer. In this embodiment, the information processing device 1 is assumed to be a server device, and for simplicity's sake will be referred to as server 1 in the following description. The server 1 acquires content such as images and text, which are the layout materials for a document page, from the terminal 2, and generates a layout image in which each content is arranged within the document page. In this embodiment, the server 1, as described below, uses a scoring model 141 (classifier) that has been trained on the layout of an existing document page through machine learning to calculate a score representing the evaluation value of the generated layout image. The server 1 returns (outputs) information about the generated layout image to the terminal 2 according to the calculated score.

Terminal 2 is a terminal device used by each user of this system, such as a personal computer, smartphone, or tablet device. Users of this system are, for example, publishers that create magazines, books, etc., and use this system to create document page layouts.

Note that the layout images generated by server 1 may not only be pages related to printed materials such as magazines and books, but may also be related to page layouts on the web, such as web pages.

Figure 2 is a block diagram showing an example configuration of server 1. Server 1 comprises a control unit 11, a main memory unit 12, a communication unit 13, and an auxiliary memory unit 14. The control unit 11 has one or more arithmetic processing devices such as a CPU (Central Processing Unit), MPU (Micro-Processing Unit), or GPU (Graphics Processing Unit), and performs various information processing, control processing, and other operations related to server 1 by reading and executing programs P stored in the auxiliary memory unit 14. The main memory unit 12 is a temporary storage area such as SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), or flash memory, and temporarily stores data necessary for the control unit 11 to execute arithmetic processing. The communication unit 13 is a communication module for performing communication-related processing, and sends and receives information to and from the outside.

The auxiliary storage unit 14 is a large-capacity memory, hard disk, etc., and stores the program P and other data necessary for the control unit 11 to execute processing. The auxiliary storage unit 14 also stores a scoring model 141 used to calculate the score of a layout image. The scoring model 141 is a trained model (classifier) used to calculate the score of a layout image, and as described below, is a trained model generated using layout images of existing document pages as training data.

The auxiliary storage unit 14 may be an external storage device connected to the server 1. The server 1 may also be a multi-computer consisting of multiple computers, or may be a virtual machine virtually constructed using software.

Furthermore, in this embodiment, the server 1 is not limited to the above configuration, and may include, for example, a reading unit that reads information stored on a portable storage medium, an input unit that accepts operation input, a display unit that displays images, etc.

Figure 3 is an explanatory diagram showing an overview of this embodiment. Figure 3 schematically shows how server 1 generates a layout image of a document page upon receiving a request from terminal 2. An overview of this embodiment will be explained based on Figure 3. First, server 1 acquires data of multiple contents to be placed on a document page from terminal 2 and accepts a request to generate a layout image. For example, server 1 acquires document data in which text, images, etc. to be placed on a document page are structured, such as an XML (Extensible Markup Language) file. Note that the content data acquired from terminal 2 is not limited to structured data and may be unstructured data.

When server 1 acquires content data from terminal 2, it also accepts input specifying the number of pages of the document to be created. As will be described later, server 1 allocates and arranges the acquired multiple pieces of content to each page, and generates a layout image for each page. The process of generating layout images for multiple pages will be described in detail later.

By executing program P, server 1 has the functions of layout information generation unit 111, layout image generation unit 112, and layout evaluation unit 113. Layout information generation unit 111 generates layout information for arranging multiple pieces of content obtained from terminal 2 within a predetermined area corresponding to a document page. Specifically, layout information generation unit 111 establishes minimum document layout specifications (aligning grids, ensuring that content does not overlap, etc.), and then randomly determines the layout coordinates for arranging each piece of content within the predetermined area, or the size of each piece of content (the size it occupies within the page), to the extent that it does not violate these specifications.

The layout information generation unit 111 randomly determines the placement coordinates, etc. of each piece of content and generates multiple patterns of layout information. Figure 3 illustrates the generation of N types of layout information. The layout information generation unit 111 generates N types of layout information with different layouts by varying the coordinates, size, etc. of each piece of content in each layout pattern.

The layout image generation unit 112 generates a layout image in which each piece of content is arranged within a predetermined area in accordance with the layout information generated by the layout information generation unit 111. Specifically, the layout image generation unit 112 generates N different layout images corresponding to the N different pieces of layout information generated by the layout information generation unit 111.

The layout evaluation unit 113 calculates a score that evaluates the likelihood of the layout for each of the N layout images generated. Specifically, the layout evaluation unit 113 calculates a score for each of the N layout images generated using a scoring model 141 (classifier) that has been trained on the layout of existing document pages. The scoring model 141 will be described in more detail below.

Based on the scores calculated above, the server 1 determines the ranking of the N layout images. The server 1 outputs information about the generated layout images to the terminal 2 according to the determined ranking. For example, the server 1 determines the top M predetermined layout images and outputs layout information (arrangement information) corresponding to the top M layout images to the terminal 2. The terminal 2 imports the layout information obtained from the server 1 into specified document editing software, and the user creates the final document page.

For example, the server 1 may output information about layout images that have a score above a threshold. Also, for example, the server 1 may output information about all generated layout images, and also present (output) the scores and rankings. In this way, the server 1 only needs to be able to present information about layout images according to the calculated scores (evaluations), and the manner in which it presents information is not limited to being based on rankings.

Figure 4 is an explanatory diagram of the layout learning process. In this embodiment, the server 1 calculates scores using a neural network constructed by deep learning, specifically a CNN (Convolution Neural Network), as the scoring model 141. Figure 4 conceptually illustrates how deep learning is performed using layout images of existing document pages as training data to construct (generate) the scoring model 141. The layout learning process for constructing the scoring model 141 will be described with reference to Figure 4.

In this embodiment, the scoring model 141 is described as being a CNN, but the scoring model 141 may also be other trained models such as other neural networks, SVMs (Support Vector Machines), Bayesian networks, decision trees, etc.

In this embodiment, the server 1 generates the scoring model 141 using a rank learning technique. Rank learning is a learning technique that learns the ranking of a dataset. Known examples of neural networks that perform rank learning using deep learning include DeepLank and SiameseNet. Because rank learning is a well-known learning technique, a detailed description of it will be omitted.

For example, server 1 acquires a manually created layout image of an existing document page from terminal 2 and uses it as training data for learning. For convenience in the following explanation, this layout image will be referred to as an "existing layout image." For example, server 1 uses the existing layout image as correct answer data with a score of "1."

The server 1 then randomly rearranges the placement coordinates of the content placed within the existing layout image, generating multiple layout images with altered content placement. For convenience in the following explanation, these layout images will be referred to as "fake layout images." The server 1 uses the fake layout images as incorrect data with a score of "0."

Server 1 inputs the existing layout image and a fake layout image in which the content arrangement has been changed from the existing layout image into a neural network and performs rank learning. Specifically, server 1 performs learning so that the score of the existing layout image will be higher than the score of the fake layout image. Server 1 compares the score of the existing layout image with the scores of each of the multiple fake layout images generated above, and performs learning so that the score of the existing layout image will be higher than the scores of any of the multiple fake layout images. In this way, server 1 generates scoring model 141. Server 1 calculates the score of the layout image using the generated scoring model 141.

In the above, it was stated that server 1 generates part of the training data (fake layout images) itself, but the training data may be created entirely by hand. Furthermore, the training data may be manually labeled with correct values such as scores and rankings.

Furthermore, the processing entity (server 1) that performs the layout learning process and the layout generation process does not have to be the same.

Figure 5 is an explanatory diagram of the layout generation process. Figure 5 conceptually illustrates the process of generating a layout image. As described above, the server 1 acquires structured data of the content to be placed on the document page from the terminal 2. Furthermore, when acquiring the content data, the server 1 also acquires the number of pages K of the document specified by the user from the terminal 2.

The server 1 allocates and arranges the multiple contents acquired from the terminal 2 to each page so that the specified number of pages is K, generating K layout images. The server 1 randomly determines the page and arrangement coordinates on which to arrange each piece of content, and generates layout images for K pages, generating a group of N layout images, as shown in Figure 5. In other words, the server 1 ultimately generates K x N layout images.

The server 1 inputs each generated layout image into a scoring model 141 and calculates a score for the layout image. In this case, the server 1 prepares a different scoring model 141 for each page of the input layout image, and inputs the layout image into the scoring model 141 corresponding to each page to calculate a score. For example, the server 1 may prepare K scoring models 141 according to the number of pages, or may prepare scoring models 141 according to the page classification within the document (for example, in the case of a magazine, the cover, table of contents page, article page, advertisement page, etc.). During learning, the server 1 performs learning using the layout image corresponding to each page as training data and generates a scoring model 141 corresponding to each page.

For each of the N layout image groups, the server 1 calculates the score of the layout image for each page using the scoring model 141 corresponding to each page. The server 1 then adds up the scores for all pages to calculate a design score. The server 1 ranks the N layout image groups based on the calculated design scores. In this way, the server 1 obtains the rankings of the generated layout images.

As already explained, the server 1 outputs information on the layout image group with the top M design scores out of the N layout image groups to the terminal 2.

As described above, the server 1 evaluates layout images using the scoring model 141 that has learned document layouts, and presents highly rated layout images to the user. This saves the user the trouble of deciding on the document layout, and improves the efficiency of document creation.

Figure 6 is a flowchart showing an example of the processing steps of the layout learning process. The processing details of the layout learning process will be described based on Figure 6. The control unit 11 of the server 1 acquires an existing layout image to be used as training data (step S11). The existing layout image is a layout image of a document page created manually. The control unit 11 generates a false layout image by rearranging each piece of content included in the existing layout image acquired in step S11 (step S12). For example, the control unit 11 randomly rearranges multiple pieces of content included in the existing layout image acquired in step S11 to generate multiple false layout images.

The control unit 11 generates a scoring model 141 that has been trained to output a score (evaluation) for a layout image when it is input, based on the existing layout image acquired in step S11 and the false layout image generated in step S12 (step S13). As described above, the control unit 11 performs training using a rank learning method to generate a scoring model 141 related to CNN. Specifically, the control unit 11 performs training using the layout image acquired in step S11 as correct data and the false layout image generated in step S12 as incorrect data. The control unit 11 then ends the series of processes.

Figure 7 is a flowchart showing an example of the processing steps of the layout generation process. The processing details of the layout generation process will be described based on Figure 7. The control unit 11 of the server 1 acquires from the terminal 2 multiple pieces of content to be arranged on document pages and the specified number of pages (step S31). The control unit 11 generates layout information that allocates and arranges multiple pieces of content to each page according to the specified number of pages (step S32). Specifically, the control unit 11 randomly determines the arrangement coordinates for arranging each piece of content within a specified area, and generates multiple patterns of layout information. The control unit 11 generates layout images of each page in which content is arranged within a specified area according to the generated layout information (step S33). Specifically, the control unit 11 generates multiple patterns of layout images according to the multiple patterns of layout information generated in step S32.

The control unit 11 calculates a score for each page of layout image generated in step S33 using a scoring model 141 that has been trained on multiple layout images (step S34). Specifically, the control unit 11 calculates a score for each page using a different scoring model 141 for each page. The control unit 11 calculates a score for each of the multiple patterns of layout images generated in step S33.

The control unit 11 adds up the scores of the layout images for each page and calculates a design score for each of the multiple patterns (step S35). The control unit 11 ranks the layout images for each pattern according to the calculated design score, and outputs layout information related to the top-ranked layout images to the terminal 2 (step S36). The control unit 11 ends this series of processes.

In the above, the scoring model 141 is assumed to output a score (evaluation value) of the layout image, but it may also be configured to output only an identification result indicating the suitability of the layout image. In other words, the server 1 only needs to be able to obtain an evaluation of the layout image using the trained model, and the obtained evaluation is not limited to a continuous probability value.

As described above, according to this first embodiment, by outputting layout image information according to the score calculated using the scoring model 141, it is possible to present an appropriate layout to the user.

Furthermore, according to this embodiment 1, the ranking of multiple layout images is determined using the scoring model 141, and the layout images are output according to the determined ranking, thereby making it possible to present a more appropriate layout to the user.

Furthermore, according to this embodiment 1, by calculating the score of a layout image using a different scoring model 141 for each page, it is possible to appropriately evaluate the layout image taking into account the characteristics of each page.

(Embodiment 2)
In this embodiment, a form will be described in which a user specifies layout conditions in advance, and a layout according to the specified conditions is output. Note that the same reference numerals will be used to designate the same parts as in the first embodiment, and a description thereof will be omitted. Fig. 8 is an explanatory diagram showing an outline of the second embodiment. The contents of Fig. 8 are almost the same as those of Fig. 3, and therefore a description of the common matters will be omitted. The outline of this embodiment will be described based on Fig. 8.

In this embodiment, when the server 1 receives a request to generate a layout image from the terminal 2, it acquires the data for each content, the specified number of pages, and condition information that specifies the conditions for arranging each content on a document page. The condition information is a layout condition arbitrarily specified by the user creating the document, and is prior knowledge information about the content that is only known to a human (user) and is necessary for determining the layout.

In this embodiment, the server 1 accepts input of condition information for images, among other content such as images and text. For example, the server 1 accepts input of the importance of each image and the order in which each image is to be arranged within a document page. The importance is input, for example, as a rank on multiple levels. The order is input, for example, in ascending numerical order.

For example, if the document is a magazine, the image size is often determined based on the importance of the image, with the image size being larger the more important it is, such as double-page spread size, full page size, or half page size. Therefore, server 1 acquires the importance of the image as condition information and determines the size of the image based on the importance.

Furthermore, when multiple images are related to each other, the order in which images should be placed first on a page is often determined based on the image content. For example, when laying out a photo of the entire subject and a photo of details on a page, it is generally better to place the photo of the entire subject first. Therefore, Server 1 acquires the order in which images should be placed as condition information.

Similar to the first embodiment, the server 1 generates a layout image in which each piece of content is randomly arranged, inputs the layout image into the scoring model 141, and calculates the design score. Furthermore, in this embodiment, the server 1 corrects the calculated design score based on the above-mentioned condition information. In other words, the server 1 changes the evaluation of the layout image in accordance with the condition information.

Specifically, the server 1 compares the size, layout order, etc. of the content (images) in the layout image of each page with the importance, layout order, etc. specified in the condition information, and determines whether each piece of content is arranged in a size, order, etc. that violates the condition information. The server 1 then counts the number of pieces of content that are arranged in a manner that violates the condition information. In other words, the server 1 counts the number of violations that violate the layout conditions specified by the user.

The server 1 multiplies the counted number of contents (number of violations) by a predetermined coefficient and subtracts the result from the design score. In other words, the server 1 imposes a penalty according to the number of violations. The server 1 ranks the images based on the design score finally calculated by the above process, and outputs information about the top layout images to the terminal 2.

As described above, the server 1 obtains condition information specifying the content placement conditions in advance and evaluates the layout image based on the condition information. This allows the layout image desired by the user to be presented.

Note that, although the condition information is used only to evaluate layout images in the above embodiment, this is not limited to this, and the server 1 may also use the condition information to generate layout images. For example, the server 1 determines the image size, etc., by referring to the importance and layout order of the images specified in the condition information, and arranges them on each page. Even in this case, the same effect as above can be achieved.

Figure 9 is a flowchart showing an example of the processing steps for layout generation processing according to embodiment 2. The control unit 11 of the server 1 acquires from the terminal 2 condition information that specifies the content placement conditions, in addition to the multiple contents to be placed on a document page and the specified number of pages (step S201). The condition information specifies the layout conditions for placing each content on a page, and is, for example, information such as the importance of the content and the placement order, as described above. The control unit 11 then proceeds to step S32.

After calculating the design score of the layout image (step S34), the control unit 11 corrects (changes) the design score based on the condition information acquired in step S201 (step S202). Specifically, the control unit 11 compares the size, arrangement order, etc. of each piece of content in the generated layout image with the importance, arrangement order, etc. of the content specified in the condition information, and determines whether each piece of content is arranged in a size, order, etc. that violates the condition information. The control unit 11 counts the number of pieces of content that are arranged in a manner that violates the condition information, and subtracts a predetermined value from the design score according to the counted number of pieces of content. The control unit 11 then proceeds to step S36.

As described above, according to this second embodiment, layout images can be evaluated according to the condition information specified by the user, making it possible to present a more appropriate layout.

Furthermore, according to this second embodiment, layout images can also be generated based on condition information specified by the user.

(Embodiment 3)
10 is a functional block diagram showing the operation of the server 1 in the above-described embodiment. The control unit 11 executes the program P, causing the server 1 to operate as follows: The acquisition unit 101 acquires multiple pieces of content. The generation unit 102 generates a layout image in which the multiple pieces of content are arranged in a predetermined area. The evaluation unit 103 uses a classifier that has trained on the multiple layout images to acquire an evaluation of the generated layout image. The output unit 104 outputs the evaluation result.

This third embodiment is as described above, and is otherwise similar to the first and second embodiments, so corresponding parts are given the same reference numerals and detailed descriptions thereof will be omitted.

(Embodiment 4)
In the above-described embodiment, an appropriate layout can be provided to the user. When determining a layout, particularly in the initial stage of consideration, it may be desirable to have a wide variety of layouts. This point will be described below.

The layout evaluation unit 113 calculates a score that evaluates the likelihood of the layout for each of the N generated layout images (as in embodiments 1 to 3), and also calculates a diversity score that represents diversity for each of the N generated layout images, and calculates an overall score based on both scores. The method for calculating the overall score is described below.

Figure 11 is an explanatory diagram showing the main parts of the scoring model 141. The scoring model 141 can be, for example, a CNN, and has fully connected layers 141a and 141b. Note that the number of fully connected layers is not limited to the example in Figure 11. In the fully connected layers 141a and 141b, both the input and output are vectors. Since the fully connected layer 141b is located before the output layer, it combines the features of the input layout image and therefore corresponds to an identification unit for classifying layouts. In embodiment 4, the vectors in the fully connected layer 141b are used as identification indices for identifying the features of the layout image. Note that other fully connected layers may also be used.

Figure 12 is an explanatory diagram showing an example of calculating the overall score. For convenience, the N layout images are represented as G1, G2, G3, ..., GN. In embodiments 1 to 3, the calculated design score is represented as STi, where i = 1 to M. For example, the design score of layout image Gi is STi. The diversity score between layout image Gi and layout image Gj is represented as SDi,j (or SD(i,j)), where j = 1 to M and i < j.

For example, in the example of Figure 12, the design score of layout image G1 is ST1, and the diversity scores between it and layout images G2, G3, ..., GN are SD(1,2), SD(1,3), ..., SD(1,N). Furthermore, the design score of layout image G2 is ST2, and the diversity scores between it and layout images G1, G3, ..., GN are SD(1,2), SD(2,3), ..., SD(2,N). Note that SD(2,1) is equal to SD(1,2), so the notation SD(1,2) is used consistently. The same applies to the other layout images.

As shown in FIG. 11, when layout image Gi out of N layout images is input to scoring model 141, the vector of fully connected layer 141b is denoted by vector di, and when layout image Gj is input to scoring model 141, the vector of fully connected layer 141b is denoted by vector dj. Here, i ≠ j. The diversity score SDi,j between layout images Gi and Gj can be calculated by subtracting the cosine similarity of vectors di and dj from 1. If the layouts of layout images Gi and Gj are similar, the cosine similarity of vectors di and dj approaches 1, and so the diversity score SDi,j approaches 0. On the other hand, if the layouts of layout images Gi and Gj are dissimilar, the cosine similarity of vectors di and dj approaches 0, and so the diversity score SDi,j approaches 1.

The overall score S can be calculated using the formula S = Σ(STi + λ·SDi,j), where Σ is the sum from 1 to N for i and j. λ is a weighting parameter (weighting coefficient) that can be set to the desired value.

Figure 13 is an explanatory diagram showing an overview of embodiment 4. The layout evaluation unit 113 functions as a calculation unit and calculates the similarity between any two layout images from among multiple layout images. The layout evaluation unit 113 calculates an overall score using a score for evaluating design and a score for evaluating diversity, and determines a selection set (subset) that maximizes the overall score. In other words, the overall score for all M layout images that make up the selection set is maximized. More specifically, the layout evaluation unit 113 can calculate an overall score by taking into account the diversity score in addition to the design score described in embodiments 1 to 3, and evaluate the layout images.

The design score STi indicates the appropriateness of the layout image, and is an index that indicates, for example, how artistic it is. On the other hand, the diversity score SDi,j is an index that indicates dissimilarity or a high degree of variation. By using the overall score S, it is possible to output layout images (top M) that are highly artistic and appropriate, and rich in diversity, as illustrated schematically in Figure 13.

Figure 14 is a flowchart showing an example of the processing steps for the layout generation processing of embodiment 4. The processing details of the layout generation processing will be described based on Figure 14. The control unit 11 of the server 1 acquires from the terminal 2 multiple pieces of content to be arranged on document pages and the specified number of pages (step S31). The control unit 11 generates layout information that allocates and arranges multiple pieces of content to each page according to the specified number of pages (step S32). Specifically, the control unit 11 randomly determines the arrangement coordinates for arranging each piece of content within a specified area, and generates multiple patterns of layout information. The control unit 11 generates layout images of each page in which the content is arranged within the specified area according to the generated layout information (step S33). Specifically, the control unit 11 generates multiple patterns of layout images according to the multiple patterns of layout information generated in step S32.

The control unit 11 calculates a score for each page of layout image generated in step S33 using a scoring model 141 that has been trained on multiple layout images (step S34). Specifically, the control unit 11 calculates a score for each page using a different scoring model 141 for each page. The control unit 11 calculates a score for each of the multiple patterns of layout images generated in step S33.

The control unit 11 adds up the scores of the layout images of each page and calculates a design score for each of the multiple patterns (step S35). The control unit 11 calculates a diversity score for the layout image of each page (S211). Specifically, the control unit 11 can calculate a diversity score for each page using a different scoring model 141 for each page. The control unit 11 calculates a diversity score for each of the multiple patterns of layout images generated in step S33.

The control unit 11 adds up the diversity scores of the layout images of each page and calculates a diversity score for each of the multiple patterns (step S212). The control unit 11 calculates an overall score for each of the multiple patterns based on the design score and diversity score (S213). The control unit 11 ranks the layout images of each pattern according to the calculated overall score, and outputs M sets (combinations) with the highest overall score to the terminal 2 (step S36). The control unit 11 ends the series of processes.

As described above, according to embodiment 4, by outputting layout image information according to the score calculated using the scoring model 141, it is possible to present to the user appropriate layouts (layouts that are highly artistic) and layouts that are rich in variety.

The embodiments disclosed herein are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the claims, not by the meaning described above, and is intended to include all modifications within the meaning and scope of the claims.

1. Server (information processing device)
11 Control unit 12 Main memory unit 13 Communication unit 14 Auxiliary memory unit P Program 141 Scoring model 141a, 141b Fully connected layer 2 Terminal

Claims (7)

an acquisition unit that acquires a plurality of contents;
a generation unit that generates a layout image in which the plurality of contents are arranged in a predetermined area;
an evaluation unit that acquires an evaluation of a generated layout image using a classifier that has been trained on a plurality of layout images;
an output unit that outputs an evaluation result, wherein the acquisition unit acquires the plurality of contents and the number of pages of the layout image,
the generating unit generates a plurality of layout images in which the plurality of contents are allocated to and arranged on each page according to the number of pages;
The information processing apparatus, wherein the evaluation unit uses a different classifier depending on the page to obtain an evaluation of each of the plurality of layout images. the generation unit generates a plurality of layout images in which the plurality of contents are arranged differently;
the evaluation unit obtains a ranking of the plurality of layout images using the classifier;
The information processing apparatus according to claim 1 , wherein the output unit outputs the ranking. the acquisition unit acquires condition information that defines a placement condition for the content;
The information processing apparatus according to claim 1 , wherein the evaluation unit changes the evaluation obtained from the classifier based on the condition information. the acquisition unit acquires condition information that defines a placement condition for the content;
4. The information processing apparatus according to claim 1, wherein the generating unit generates the layout image based on the condition information. a calculation unit that calculates a similarity between any two layout images among the plurality of layout images;
5. The information processing apparatus according to claim 1, wherein the evaluation unit evaluates the plurality of layout images using the evaluation value by the classifier and the similarity calculated by the calculation unit. Acquire multiple contents and the number of pages of the layout image,
generating a plurality of layout images in which the plurality of contents are allocated to each page according to the number of pages and arranged in a predetermined area;
acquiring an evaluation of each of the plurality of layout images generated using different trained classifiers according to the pages;
An information processing method comprising causing a computer to execute a process of outputting an evaluation result. Acquire multiple contents and the number of pages of the layout image,
generating a plurality of layout images in which the plurality of contents are allocated to each page according to the number of pages and arranged in a predetermined area;
acquiring an evaluation of each of the plurality of layout images generated using different trained classifiers according to the pages;
A program that causes a computer to execute a process of outputting an evaluation result.