Information processing apparatus, printing setting prompting method, and program product

Abstract

This invention relates to an information processing apparatus, a printing setting prompting method, and a program product, enabling users to easily select appropriate printing settings when printing on fabric. The information processing apparatus includes: a registration data storage unit that, for each of a plurality of fabric images, stores registration data corresponding to a vector (i.e., an embedding representation) representing the surface characteristics of the fabric and determined based on the fabric image, and the type of fabric (i.e., fabric variety) represented by the fabric image; a printing setting data storage unit that, for each of a plurality of fabric varieties, stores printing setting data corresponding to the printing settings of the printing apparatus; a fabric variety determination unit that determines a fabric variety that is the same as or similar to the target fabric by comparing the object embedding representation of an image of the target fabric obtained from photographing the target fabric with the registration embedding representation contained in the registration data; and a prompting unit that prompts the user to establish a corresponding printing setting in the printing setting data for the same or similar fabric variety determined for the target fabric.

Description

Technical Field

[0001] This disclosure relates to information processing apparatus, printing setting prompts, and procedures. Background Technology

[0002] In the technology described in Patent Document 1, the feature values ​​of data that have been printed in the past are pre-registered in correspondence with the printing settings of the printing apparatus, and the feature values ​​of data that will be printed during printing are recommended to be corresponding with similar feature values ​​in the printing settings.

[0003] Patent Document 1: Japanese Patent Application Publication No. 2020-30594

[0004] A printing press for printing inks onto fabric is known. In printing, even when printing the same image, it is sometimes necessary to change printing settings such as the ink application rate depending on the type of fabric being printed. Compared to printing on paper, when using fabric as the printing medium, the cost tends to increase with the number of reprints. Therefore, it is desirable to have a technology that allows users to easily select appropriate printing settings corresponding to the type of fabric. Summary of the Invention

[0005] This disclosure can be implemented in the following ways.

[0006] According to a first aspect of this disclosure, an information processing apparatus is provided to prompt printing settings for a printing apparatus using fabric as the printing medium. This information processing apparatus includes: a registration data storage unit that, for each of a plurality of fabric images obtained by photographing a plurality of fabrics, stores registration data that establishes a correspondence between a vector representing the surface characteristics of the fabric, i.e., an embedding, determined based on the fabric image, and the type of fabric represented by the fabric image, i.e., a fabric type; a printing setting data storage unit that, for each of a plurality of fabric types, stores printing setting data that establishes a correspondence between the fabric type and the printing settings of the printing apparatus; a fabric type determination unit that determines a fabric type that is the same as or similar to the target fabric by comparing the embedding, i.e., an object embedding, of the target fabric image obtained by photographing the target fabric with the embedding, i.e., a registration embedding, contained in the registration data; and a prompting unit that, regarding the same or similar fabric type determined for the target fabric, prompts the establishment of the corresponding printing settings in the printing setting data.

[0007] According to a second aspect of this disclosure, a method for prompting printing settings for a printing apparatus using fabric as the printing medium is provided. This method includes: for each of a plurality of fabric images obtained by photographing a plurality of fabrics, a step of preparing registration data by establishing a correspondence between a vector representing the surface characteristics of the fabric, i.e., an embedding, determined based on the fabric image, and the type of fabric represented by the fabric image; for each of the plurality of fabric types, a step of preparing printing setting data that establishes a correspondence between the fabric type and the printing settings of the printing apparatus; a step of determining a fabric type that is the same as or similar to the target fabric by comparing the embedding, i.e., the object embedding, of the target fabric image obtained by photographing the target fabric with the embedding, i.e., the registration embedding, contained in the registration data; and a step of prompting the establishment of a corresponding printing setting in the printing setting data regarding the same or similar fabric type determined for the target fabric.

[0008] According to a third aspect of this disclosure, a program executed by a computer is provided to prompt printing settings for a printing apparatus using cloth as the printing medium. This program enables the computer to: for each of a plurality of cloth images obtained by photographing a plurality of cloths, prepare registration data corresponding to an embedding representation (vector) representing the surface characteristics of the cloth and determined based on the cloth image, and the type of cloth represented by the cloth image; for each of the plurality of cloth types, prepare printing setting data corresponding to the cloth type and the printing settings of the printing apparatus; determine a cloth type that is the same as or similar to the target cloth by comparing the embedding representation (object embedding) of the target cloth image obtained by photographing the target cloth with the embedding representation (registered embedding) contained in the registration data; and prompt the computer to establish the corresponding printing settings in the printing setting data regarding the same or similar cloth type determined for the target cloth. Attached Figure Description

[0009] Figure 1 This is an explanatory diagram showing the general configuration of the printing system according to this embodiment.

[0010] Figure 2 This is an explanatory diagram showing an example of printing setting data.

[0011] Figure 3 This is an illustrative diagram about creating segmented images.

[0012] Figure 4 This is a flowchart illustrating the process of registering registration data.

[0013] Figure 5 This is an explanatory diagram showing an example of registered data.

[0014] Figure 6 This is the first half of a flowchart showing the processes involved in the printing settings prompts.

[0015] Figure 7 This is the latter half of the flowchart showing the processes involved in the printing settings prompts.

[0016] Figure 8 This is an explanatory diagram showing an example of data used for comparison.

[0017] Figure 9 This is an illustrative diagram showing an example of distance data.

[0018] Explanation of reference numerals in the attached figures

[0019] 10… Printing system, 100… Information processing device, 101… Memory, 102… Interface circuit, 103… Input device, 104… Display device, 105… Processor, 110… Embedded conversion unit, 120… Fabric type determination unit, 130… Prompt unit, 140… Learning unit, 150… Update unit, 200… Camera, 300… Printing apparatus, Cr… Segmented image, D1… Registration data, D2… Printing setting data, D3… Comparison data, D4… Distance data, FB… Fabric, IMfb… Image, M1… Conversion model, P1… Program, St… Stride, i, j… Index. Detailed Implementation

[0020] A. Implementation method: Figure 1 This is an explanatory diagram showing the general configuration of the printing system 10 according to this embodiment. The printing system 10 includes an information processing device 100, a camera 200, and a printing apparatus 300. The printing apparatus 300 is a printing press that performs printing on fabric printing media.

[0021] The information processing device 100 prompts the printing device 300, which performs printing on a fabric printing medium, with appropriate printing settings. The information processing device 100 may be, for example, a personal computer. The information processing device 100 includes a memory 101, an interface circuit 102, an input device 103 connected to the interface circuit 102, a display device 104, and a processor 105. The memory 101 stores programs and data used in various processes performed by the information processing device 100. In the illustrated example, the memory 101 stores program P1, registration data D1, printing setting data D2, and conversion model M1. Registration data D1, printing setting data D2, and conversion model M1 will be described later.

[0022] Camera 200 is used to photograph printing media. For example, camera 200 photographs fabric FB, which is a printing medium arranged on a mounting table (not shown), from directly above. Camera 200 outputs the fabric image obtained by photographing the printing medium to information processing device 100. The fabric image is an image obtained by photographing the fabric that has become a printing medium.

[0023] The information processing device 100 includes an embedded conversion unit 110, a seed determination unit 120, a prompting unit 130, a learning unit 140, and an update unit 150. The processor 105 executes the program P1 stored in the memory 101, thereby realizing the functions of these units.

[0024] The embedding conversion unit 110 uses the conversion model M1, described later, to convert the fabric image into an embedding representation. The embedding representation represents the surface properties of the fabric. The embedding representation is a vector of values ​​converted from the fabric image.

[0025] The fabric type determination unit 120 determines the same or similar fabric type as the target fabric by comparing the embedded representation of the target fabric image with the embedded representation contained in the registration data D1. "Target fabric" refers to the fabric of the printing medium that is the object of the prompt set for printing. "Target fabric image" is a fabric image obtained by photographing the target fabric. Fabric types include, for example, wide-width fabric, satin, chiffon, and nylon.

[0026] Registration data D1 is data that establishes a correspondence between the embedded representation converted from the fabric image and the fabric identification value that identifies the type of fabric represented by the fabric image for each of the multiple fabric images obtained by photographing multiple fabrics. Registration data D1 is created by update unit 150. The memory 101 that stores registration data D1 is also called "registration data storage unit".

[0027] The prompting unit 130 prompts the corresponding printing settings to be created in the printing setting data D2 regarding the fabric type determined by the fabric type determination unit 120.

[0028] Figure 2 This is an explanatory diagram showing an example of print setting data D2. Print setting data D2 is data that establishes a correspondence between the fabric identification value (where M is a positive integer) and the print settings of the printing apparatus 300 for each of M types of fabric. The memory 101 storing the print setting data D2 is also called a "print setting data storage unit". The print settings include settings for image resolution, total ink volume, and number of cycles. The number of cycles indicates the number of times the printing head reciprocates in order to print within a determined printing range. The print setting data D2 is pre-created by the information processing apparatus 100 or by a device other than the information processing apparatus 100.

[0029] Learning Unit 140 creates a transformation model M1 that converts a cloth image into an embedded representation by performing machine learning.

[0030] For each of the multiple fabric images, the update unit 150 creates registration data D1 corresponding to the fabric identification value and the embedded representation converted from the fabric image. Additionally, the update unit 150 may, for example, change the content of records contained in the registration data D1 in response to a user's instruction. The update unit 150 may, for example, add new records to the registration data D1 in response to a user's instruction. The update unit 150 may, for example, delete a portion of the records contained in the registration data D1 in response to a user's instruction. The update unit 150 may also be referred to as a "registration data creation unit".

[0031] In this embodiment, a learning dataset for use in machine learning to create the transformation model M1 is prepared as follows. For example, the learning unit 140 prepares the learning dataset. The preparation of the learning dataset includes the creation of segmented images. First, the camera 200 photographs M types of cloth (M is a positive integer), obtaining M cloth images. One of the M cloth images is segmented into N (N is a positive integer), thereby creating N segmented images.

[0032] Figure 3 This is an illustrative diagram about creating a segmented image. Figure 3 The image shown is an example of creating segmented images after segmenting the image IMfb. The image IMfb is a photograph of the fabric. Here, the fabric was photographed with the background not included in the image IMfb. A pre-defined square selection area is chosen from the image IMfb. The selected square area constitutes one segmented image Cr. Figure 3 In the diagram, a selected region representing one segmented image Cr is indicated by a dashed box. The selected region slides at preset steps St. In the illustrated example, two consecutive selected regions in the X-axis direction have overlapping portions. The illustrated example shows the sliding of the selected region along the X-axis direction, but the selected region can also slide sequentially along the X-axis and Y-axis directions. Furthermore, two consecutive selected regions in the Y-axis direction may also have overlapping portions. Alternatively, two consecutive selected regions in the X-axis direction may not have overlapping portions. Two consecutive selected regions in the Y-axis direction may also not have overlapping portions.

[0033] The fabric type, represented by the unsegmented fabric image, is used as a label to establish a correspondence between each of the N segmented images. The same fabric type is used as a label to establish a correspondence between N segmented images created from the same fabric image.

[0034] Furthermore, the M cloth images are each segmented into N segments. The cloth type is used as a label to establish a correspondence between each segmented image. This process prepares an M×N set of segmented images. The set of segmented images and label groups is used as the training dataset.

[0035] The following describes the machine learning performed by the learning unit 140 in order to create the transformation model M1. As a first step, the learning unit 140 performs self-supervised training. For example, SimSiam (Xinlei Chen, 1st author, "SimSiam: Exploring Simple Siamese Representation Learning", [online], [accessed March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.2011.10566) can be used as a method to transform the embedded representations obtained by performing data augmentation on the learning cloth image and the learning image, respectively, so that their embedding representations are close to each other in the embedding space. Data augmentation includes rotation, inversion, cropping, tone conversion, etc. Since self-supervised training is performed in the first step, the segmented images of the learning dataset are used as input data, but the labels of the learning dataset are not used. When a machine learning model created through self-supervised learning is input to two or more cloth images, the machine learning model outputs an embedding representation transformed from each cloth image.

[0036] As a second step, the learning unit 140 performs transfer learning on the machine learning model created by performing self-supervised learning. In this embodiment, the learning unit 140 performs deep distance learning in the transfer learning. The learning unit 140 adjusts only the output layer while freezing the output layer of the machine learning model created by performing self-supervised learning. Freezing the output layer means that the parameters outside the output layer are used as is. As a method for deep distance learning, for example, FaceNet (Florian Schroff, 2 others, "FaceNet: A Unified Embedding for FaceRecognition and Clustering", [online], [accessed March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.1503.03832). In the second step of deep distance learning, referring to the fabric types indicated by the labels in the learning dataset, learning is performed such that the embedding representations transformed from two or more learning fabric images depicting the same fabric type are close to each other in the embedding space, while the embedding representations transformed from two or more learning fabric images depicting different fabric types are far apart in the embedding space. If the values ​​corresponding to the labels of two or more learning fabric images are the same, the fabric types represented by the two or more learning fabric images are the same. If the values ​​corresponding to the labels of two or more learning fabric images are different, the fabric types represented by the two or more learning fabric images are different. The transformation model M1 created through the execution of deep distance learning outputs the embedding representations transformed from the two or more learning fabric images.

[0037] In this embodiment, learning is performed without labels in the self-supervised learning of the first step, so efficient learning can be achieved even when using a large-scale learning dataset.

[0038] Figure 4 This is a flowchart illustrating the process of registering registration data D1. The transformation model M1 is... Figure 4 The start time of the processing is generated by the method described above.

[0039] In step S101, multiple segmentation images are prepared. Here, the M×N segmentation images contained in the learning dataset are used.

[0040] In step S102, the embedding conversion unit 110 inputs one segmented image into the conversion model M1 for the i-th seed (i is a positive integer and a value less than or equal to M), and obtains the embedding representation output from the conversion model M1. i is the index used to distinguish the seed. The initial value of index i is 1.

[0041] The embedding conversion unit 110 stores the cloth recognition value, the image number of the recognized segmented image, and the record corresponding to the acquired embedding representation in the memory 101. In step S102, the set of records stored in the memory 101 constitutes the registration data D1.

[0042] Figure 5 This is an explanatory diagram illustrating an example of registration data D1. The cloth identification value is a value representing the type of cloth, as indicated by the segmented image used as the transformation source and embedded in the segmented image input to transformation model M1. The image number is the value used to identify the segmented image. In the illustrated example, multiple segmented images segmented from the same cloth image are assigned sequence numbers. In this embodiment, the embedded representation is represented by a vector consisting of 512 numerical values ​​arranged side-by-side. That is, the embedded representation is 512-dimensional data. In the illustrated example, each embedded representation contained in set G1 is transformed from each of the N segmented images segmented from the same cloth image. The embedded representation contained in registration data D1 is also referred to as the "registration embedded representation".

[0043] like Figure 4 As shown, in step S103, the update unit 150 determines whether the embedding representation based on each of the N segmented images has been registered for the i-th fabric type. If the update unit 150 determines that the embedding representation has been registered for all N segmented images for the i-th fabric type (step S103; "Yes"), the processing of step S104 is executed. On the other hand, if the update unit 150 determines that the embedding representation based on each of the N segmented images has not been registered for the i-th fabric type (step S103; "No"), the processing of step S102 is executed again.

[0044] In step S104, the update unit 150 determines whether the registration of embedded representations has been completed for all M fabric types. If the update unit 150 has not completed the registration of embedded representations for all M fabric types (step S104; "No"), the process in step S105 is executed. If the update unit 150 has completed the registration of embedded representations for all M fabric types (step S104; "Yes"), the process ends. Figure 4 The processing shown.

[0045] In step S105, the update unit 150 increments the index i. Then, the process of step S102 is executed again.

[0046] Figures 6-7 This is a flowchart illustrating the processes involved in determining the type of fabric and setting the printing parameters. Figure 6 The start time of processing, the registration processing of registration data D1 (refer to...) Figure 4 This is complete. That is, registration data D1 has been registered.

[0047] In step S201, the fabric type determination unit 120 creates N segmented images for the target fabric image. First, the fabric type determination unit 120 controls the camera 200 to capture an image of the target fabric. Assume the background of the target fabric image does not contain any fabric. The fabric type determination unit 120 segments the target fabric image into N segments, creating N segmented images. The method for segmenting the fabric image is the same as when preparing the learning dataset (see [reference]). Figure 3 ).

[0048] In step S202, the embedding conversion unit 110 inputs one of the N segmented images into the conversion model M1 and obtains the embedding representation output from the conversion model M1. In this embodiment, based on one segmented image, the embedding representation is obtained as a vector composed of 512 values ​​arranged side by side. The embedding representation based on the input segmented image is also referred to as the "object embedding representation".

[0049] Figure 8 This is an explanatory diagram showing an example of the comparison data D3.

[0050] The embedding conversion unit 110 establishes corresponding records for the image number of the identified segmented image and the acquired embedding representation and saves them to the memory 101. In step S203, the set of records saved in the memory 101 constitutes the comparison data D3. Furthermore, the comparison data D3 is data temporarily created during the prompt processing of the printing settings, therefore... Figure 1 The control data D3 is not shown in the figure.

[0051] like Figure 6 As shown, in step S203, the seeding determination unit 120 determines whether it has acquired embedding representations for all N segmented images. If the seeding determination unit 120 determines that embedding representations have been acquired for all N segmented images (step S203; "Yes"), the processing in step S204 is executed. On the other hand, if the seeding determination unit 120 determines that it has not acquired embedding representations for all N segmented images (step S203; "No"), the processing in step S202 is executed again.

[0052] In step S204, the fabric type determination unit 120 calculates index values ​​representing the proximity of the N vectors corresponding to the object fabric included in the control data D3 and the N vectors corresponding to the j-th fabric included in the registration data D1. j is an index used to distinguish the fabric type. In this embodiment, index j is consistent with the fabric identification value. Index j is a positive integer less than or equal to M. The initial value of index j is 1.

[0053] As indicators of the closeness between two vectors, Euclidean distance and cosine similarity can be used, for example. Euclidean distance can be calculated by taking the square root of the sum of the squared differences between the components of the two vectors. Cosine similarity can be calculated by dividing the inner product of the two vectors by the product of their Euclidean norms. Here, we will illustrate an example of using Euclidean distance as an indicator of the closeness between two vectors.

[0054] If the Euclidean distance between each of the N vectors contained in the control data D3 and each of the N vectors of a single fabric type contained in the registered data D1 is calculated in the fabric type determination unit 120, then N can be obtained. 2 The Euclidean distance. The smaller the Euclidean distance value, the closer the two vectors are. (The following appears to be a separate, unrelated section:) Selection of N in the seed determination section 120. 2 N in ascending order of Euclidean distances 2 / 2 Euclidean distance. The cloth type determination unit 120 calculates the selected N. 2 The average of / 2 Euclidean distances. In this embodiment, the calculated average is set as the distance between the embedded representation of the object cloth and the embedded representation of a cloth identification value contained in the registration data D1.

[0055] Figure 9 This is an explanatory diagram showing an example of distance data D4, which represents the distance between the embedded representation of the target fabric and the embedded representation included in the registration data D1. The fabric type determination unit 120 establishes a corresponding record of the fabric identification value and the calculated average value and saves it to the memory 101. The set of records saved in the memory 101 constitutes the distance data D4. Furthermore, the distance data D4 is temporary data created during the printing setting prompt processing, therefore... Figure 1 The distance data D4 is not shown in the figure.

[0056] like Figure 6 As shown, in step S205, the fabric type determination unit 120 determines whether it has calculated the distance between the embedding representation of the target fabric and the embedding representation contained in the registration data D1 for all M fabric types. If the fabric type determination unit 120 determines that it has not calculated the distance between the embedding representation of the target fabric and the embedding representation contained in the registration data D1 for all M fabric types (step S205; "No"), the process of step S206 is executed. If the fabric type determination unit 120 determines that it has calculated the distance between the embedding representation of the target fabric and the embedding representation contained in the registration data D1 for all M fabric types (step S205; "Yes"), the process of step S207 is executed.

[0057] In step S206, the seed determination unit 120 increments the index j. Then, the process of step S204 is executed again.

[0058] like Figure 7 As shown, in step S207, the fabric type determination unit 120 determines a fabric type that is the same as or similar to the target fabric. First, the fabric type determination unit 120 determines the distance data D4 (refer to...). Figure 9 The fabric type determination unit 120 selects the minimum value among the multiple values ​​if there are multiple values ​​below the first threshold among the distances included in the distance data D4. The fabric type determination unit 120 determines a fabric identification value in the distance data D4 that corresponds to a distance below the first threshold. The fabric type determination unit 120 determines that the fabric type of the determined fabric identification value is the same as the fabric type of the target fabric. Furthermore, "same as the fabric type of the target fabric" here means not only strictly the same as the target fabric type, but also that there is no substantial difference between the fabric type and the target fabric type, and it is considered the same as the target fabric type. If there are no values ​​below the first threshold among the distances included in the distance data D4, the fabric type determination unit 120 determines a value below a preset second threshold among the distances included in the distance data D4. The second threshold is set to a value exceeding the first threshold. The fabric type determination unit 120 determines a fabric identification value in the distance data D4 that corresponds to a distance below the second threshold. The fabric type determination unit 120 determines that the fabric type showing the determined fabric identification value is similar to the fabric type of the target fabric.

[0059] In step S208, the prompting unit 130 retrieves the printing setting data D2 (refer to...) Figure 2 Obtain the printing settings corresponding to the same or similar fabric types determined in step S207.

[0060] In step S209, the prompting unit 130 outputs an image indicating the printing settings to the display device 104. The display device 104 displays the image indicating the printing settings (not shown).

[0061] As explained above, in this embodiment, by comparing the embedded representation of the target fabric image with the embedded representation contained in the registration data, a fabric type that is the same as or similar to the target fabric can be determined. Furthermore, a prompt is made to establish a corresponding printing setting for the determined fabric type. In printing, even when printing the same image, there is sometimes a need to change printing settings such as the ink application amount depending on the type of fabric used as the printing medium. Compared to the case where the printing medium is paper, there is a tendency for the cost to increase with the number of reprints when using fabric as the printing medium. In this embodiment, since suitable printing settings corresponding to the fabric type are prompted, the user can easily select the printing setting when printing on the target fabric. This helps to suppress the cost increase caused by the increase in reprints.

[0062] B. Other implementation methods: (B1) In the above embodiment, in order to determine the same type of fabric as the target fabric, the fabric type determination unit 120 first determines a value less than a first threshold among the distances included in the distance data D4. If there is no distance less than the first threshold among the distances in the distance data D4, the fabric type determination unit 120 determines a value less than or equal to a second threshold among the distances included in the distance data D4 (see reference). Figure 6 Step S207).

[0063] Alternatively, regardless of whether there is a distance in the distance data D4 that is less than the first threshold, the cloth determination unit 120 determines the value below the second threshold among the distances contained in the distance data D4.

[0064] (B2) Alternatively, if the fabric type determination unit 120 determines multiple fabric types as the same or similar fabric types for the target fabric, the prompting unit 130 provides printing settings corresponding to each of the determined multiple fabric types. When the user performs printing on the target fabric used as the printing medium, they can select the desired printing settings from the appropriate printing settings.

[0065] Alternatively, if the fabric selection unit 120 selects multiple fabric types as the same or similar to the target fabric, it may prompt the user with the corresponding printing settings for a predetermined number of fabric types, arranged in order of proximity to the target fabric. When printing on the target fabric used as the printing medium, the user can select the desired printing settings from among the more suitable options.

[0066] Alternatively, the prompting unit 130 may prompt the user along with a representative fabric image representing the same or similar fabric type as the target fabric, establishing corresponding printing settings. For example, one of the N segmented images Cr used when creating registration data D1 can be used as the representative fabric image. In this case, for example, a fabric image table corresponding to the fabric identification value and the storage destination address of the segmented image Cr serving as the representative fabric image is pre-stored in memory 101. Additionally, the representative fabric image is pre-stored in memory 101. The prompting unit 130 determines the storage destination address of the representative fabric image based on the fabric identification value of the same or similar fabric type as the target fabric, referring to the fabric image table. The prompting unit 130 outputs the printing settings, including the representative fabric image stored at the determined storage destination address, to the display device 104.

[0067] (B3) In the above embodiments, an example was described in which the information processing device 100 determines a type of fabric that is the same as or similar to the target fabric and prompts for establishing corresponding printing settings for the determined same or similar fabric. This disclosure is not limited to the scenario of prompting printing settings, but can also be applied in scenarios of prompting settings related to preprocessing and drying.

[0068] Printed materials, such as headscarves, display the same pattern on both the front and back of a fabric. When creating such printed materials using dyeing and printing, a pretreatment is sometimes performed before printing to impregnate the fabric with a penetrant that guides the ink. The settings involved in the pretreatment include, for example, the type of penetrant, the concentration of the penetrant, and the impregnation time. When the pretreatment settings are suggested based on the type of fabric, data corresponding to the fabric identification value and pretreatment settings for each of multiple fabric types is prepared in advance. The information processing device 100 can determine a fabric type that is the same as or similar to the target fabric and suggest pretreatment settings corresponding to the determined same or similar fabric type. Since candidates for suitable pretreatment settings are suggested based on the type of fabric of the printing medium, the user can easily select the pretreatment settings.

[0069] Drying is the process of drying the printing medium using a dryer after printing. The settings related to drying include temperature and drying time. When the drying settings are prompted based on the fabric type, data corresponding to the fabric identification value and drying settings is prepared in advance for each of multiple fabric types. The information processing device 100 can determine a fabric type that is the same as or similar to the target fabric and prompts for the corresponding drying settings. Because it prompts for candidates of suitable drying settings corresponding to the fabric type, the user can easily select the appropriate drying settings.

[0070] (B4) In the above implementation, an example of using SimSiam as a self-supervised learning method in the first step of creating the transformation model M1 was described. Alternatively, any of the following can be used as a self-supervised learning method.

[0071] SimCLR (Ting Chen, 3 others, "SimCLR: A Simple Framework for Contrastive Learning of Visual Representations", [online], [accessed March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.2002.05709) BYOL (Jean-Bastien Grill, 13 others, "BYOL: Bootstrap your own latent", [online], [accessed March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.2006.07733)

[0072] SwAV: Swapping Assignments between Views (Mathilde Caron, 5 others, “Unsupervised Learning of Visual Features by Contrasting ClusterAssignments”, [online], [retrieved March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.2006.09882)

[0073] In the above implementation, an example of using FaceNet as a method for deep distance learning in the second step of creating the transformation model M1 was described. Alternatively, any of the following can be used as a method for deep distance learning.

[0074] CosFace (Hao Wang, 7 others, “CosFace: Large Margin Cosine Loss for DeepFace Recognition”, [online], [retrieved March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.1801.09414)

[0075] ArcFace (Jiankang Deng, 5 others, "ArcFace: Additive Angular Margin Loss for Deep Face Recognition", [online], [retrieved March 20, 2025], Internet URL: https: / / doi.org / 10.48550 / arXiv.1801.07698)

[0076] In the above implementation, an example was described of performing self-supervised learning in the first step and deep distance learning as transfer learning in the second step to create a conversion model M1. Alternatively, the conversion model M1 can also be created by performing only deep distance learning. In deep distance learning, the learning is performed such that the embedding representations transformed from two or more learning images of fabrics of the same type are close to each other in the embedding space, and the embedding representations transformed from two or more learning images of fabrics of different types are far apart in the embedding space.

[0077] This disclosure is not limited to the embodiments described above, and can be implemented with various configurations without departing from its spirit. For example, technical features in embodiments corresponding to the technical features in the various methods described in the invention's headings can be appropriately replaced or combined to solve some or all of the above-described problems, or to achieve some or all of the above-described effects. Furthermore, if a technical feature is not described as essential in this specification, it can be appropriately omitted.

[0078] C. Other methods: (1) According to a first aspect of the present disclosure, an information processing apparatus is provided to prompt printing settings of a printing apparatus using cloth as the printing medium. The information processing apparatus includes: a registration data storage unit that, for each of a plurality of cloth images obtained by photographing a plurality of cloths, stores registration data that establishes a correspondence between a vector representing the surface characteristics of the cloth, i.e., an embedding, determined based on the cloth image, and the type of cloth represented by the cloth image, i.e., a cloth type; a printing setting data storage unit that, for each of a plurality of cloth types, stores printing setting data that establishes a correspondence between the cloth type and the printing settings of the printing apparatus; a cloth type determination unit that determines a cloth type that is the same as or similar to the target cloth by comparing the embedding, i.e., an object embedding, of the target cloth image obtained by photographing the target cloth with the embedding, i.e., a registration embedding, contained in the registration data; and a prompting unit that, regarding the same or similar cloth type determined for the target cloth, prompts the establishment of the corresponding printing settings in the printing setting data.

[0079] Based on the above method, the system prompts users to establish corresponding printing settings for fabrics that are the same as or similar to the target fabric used as the printing medium. Therefore, users can easily select printing settings when printing on the target fabric.

[0080] (2) In the information processing apparatus described above, the fabric type determination unit may determine the same or similar fabric type for the object fabric based on the proximity between the object embedding representation and the registered embedding representation in the embedding space of the embedding representation during the comparison.

[0081] (3) In the information processing device described above, the fabric type determination unit may determine that the fabric type corresponding to the registered embedding representation whose index value representing the proximity between the object embedding representation and the registered embedding representation is less than a predetermined first threshold is the same as the object fabric type.

[0082] (4) In the information processing device described above, the fabric type determination unit may determine that the fabric type corresponding to the registration embedding representation whose index value is less than a second threshold predetermined as exceeding the first threshold is a fabric type similar to the object fabric.

[0083] (5) In the information processing device described above, it is also possible that when the fabric type determination unit determines multiple identical or similar fabric types for the target fabric, the prompting unit sets the printing settings corresponding to each of the multiple identical or similar fabric types determined for the target fabric.

[0084] Users can choose their desired print settings from a number of suggested options.

[0085] (6) In the information processing apparatus described above, an embedding conversion unit may also be included, which uses a machine learning model to convert the object fabric image into the embedding representation. Alternatively, the machine learning model may be generated by performing deep distance learning, where the embedding representations converted from two or more learning fabric images of the same type of fabric are close to each other in the embedding space, and the embedding representations converted from two or more learning fabric images of different types of fabric are far apart in the embedding space.

[0086] (7) In the information processing device described above, there may also be a registration data creation unit, which acquires multiple fabric images by dividing the image obtained by photographing the fabric into N (N is a positive integer) for each of the multiple fabric images, and creates the registration data by establishing a correspondence between the embedded representation converted by the embedding conversion unit and the fabric type for each of the multiple fabric images.

[0087] (8) In the information processing device described above, the prompting unit may also prompt a representative fabric image of the same or similar fabric type determined together with the printing settings corresponding to the same or similar fabric type determined for the target fabric.

[0088] (9) According to a second aspect of this disclosure, a method for prompting printing settings for a printing apparatus using cloth as the printing medium is provided. This method includes: for each of a plurality of cloth images obtained by photographing a plurality of cloths, a step of preparing registration data by establishing a correspondence between a vector representing the surface characteristics of the cloth, i.e., an embedding, determined based on the cloth image, and the type of cloth represented by the cloth image; for each of the plurality of cloth types, a step of preparing printing setting data that establishes a correspondence between the cloth type and the printing settings of the printing apparatus; a step of determining a cloth type that is the same as or similar to the target cloth by comparing the embedding, i.e., the target embedding, of the target cloth image obtained by photographing the target cloth with the embedding, i.e., the registration embedding, contained in the registration data; and a step of prompting the establishment of a corresponding printing setting in the printing setting data regarding the same or similar cloth type determined for the target cloth.

[0089] (10) According to a third aspect of this disclosure, a program executed by a computer is provided to prompt the printing settings of a printing apparatus using cloth as the printing medium. This program enables the computer to: for each of a plurality of cloth images obtained by photographing a plurality of cloths, prepare registration data corresponding to an embedding representation (vector) representing the surface characteristics of the cloth and determined based on the cloth image, and the type of cloth represented by the cloth image; for each of the plurality of cloth types, prepare printing setting data corresponding to the cloth type and the printing settings of the printing apparatus; determine a cloth type that is the same as or similar to the object cloth by comparing the embedding representation (object embedding) of the object cloth image obtained by photographing the object cloth with the embedding representation (registered embedding) contained in the registration data; and prompt the computer to establish the corresponding printing settings in the printing setting data regarding the same or similar cloth type determined for the object cloth.

Claims

1. An information processing device that provides prompts for printing settings of a printing apparatus using cloth as the printing medium, characterized in that, have: The registration data storage unit stores registration data for each of the multiple fabric images obtained by photographing multiple fabrics. In the registration data, a vector representing the surface characteristics of the fabric and determined based on the fabric image is embedded to establish a correspondence with the type of fabric represented by the fabric image. The printing setting data storage unit stores printing setting data that establishes a correspondence between the printing setting of the printing device and the fabric for each of the multiple fabric types. The fabric type determination unit determines fabric types that are the same as or similar to the target fabric by comparing the embedding representation of the target fabric image obtained based on a photograph of the target fabric, i.e., the object embedding representation, with the embedding representation contained in the registration data, i.e., the registration embedding representation; and The prompt section indicates that, regarding the same or similar fabric type determined for the target fabric, a corresponding printing setting should be established in the printing setting data.

2. The information processing apparatus according to claim 1, wherein, The fabric type determination section In the comparison, based on the proximity of the object embedding representation within the embedding space of the embedding representation to the registered embedding representation, the same or similar fabric type is determined for the object fabric.

3. The information processing apparatus according to claim 2, wherein, The fabric type determination section If the index value representing the proximity between the object embedding representation and the registered embedding representation is less than a predetermined first threshold, the corresponding fabric type of the registered embedding representation is the same as that of the object embedding representation.

4. The information processing apparatus according to claim 3, wherein, The fabric type determination section The registration embedding that determines the value of the index to be less than a second threshold that is predetermined as a value exceeding the first threshold indicates that the corresponding fabric type is a fabric type similar to the object fabric.

5. The information processing apparatus according to claim 4, wherein, When the fabric type determination unit determines multiple identical or similar fabric types for the target fabric, The prompt section specifies the printing settings for each of the multiple identical or similar fabric types determined for the target fabric.

6. The information processing apparatus according to claim 5, wherein, The information processing device further includes an embedding conversion unit that uses a machine learning model to convert the object layout image into the embedding representation. Machine learning models It is generated by performing deep distance learning, which involves converting the embedding representations of two or more learning fabric images of the same type of fabric into each other in the embedding space, and converting the embedding representations of two or more learning fabric images of different types of fabric into each other in the embedding space.

7. The information processing apparatus according to claim 6, wherein, The information processing device also includes a registration data creation unit. The registration data creation unit acquires multiple fabric images for each of several different types of fabric by dividing the image obtained from photographing the fabric into N images, where N is a positive integer. For each of the plurality of fabric images, the registration data is created by establishing a correspondence between the embedded representation converted by the embedding conversion unit and the fabric type.

8. The information processing apparatus according to any one of claims 1-7, wherein, The prompt section, together with the printing settings corresponding to the same or similar fabric types determined for the target fabric, indicates a representative fabric image of the determined same or similar fabric types.

9. A method for providing printing setting prompts, characterized in that, This is a printing setting prompting method for printing devices that use cloth as the printing medium, comprising the following steps: For each of the multiple fabric images obtained by photographing multiple fabrics, registration data is prepared by establishing a correspondence between the vector representing the surface characteristics of the fabric and the embedded representation determined based on the fabric image and the type of fabric represented by the fabric image. For each of the multiple fabric types, corresponding printing setting data is prepared to be established between the fabric type and the printing settings of the printing device; By comparing the embedding representation of the object fabric image obtained based on the photographed object fabric (i.e., the object embedding representation) with the embedding representation contained in the registration data (i.e., the registration embedding representation), the fabric type that is the same as or similar to the object fabric is determined; and Regarding the same or similar fabric types determined for the target fabric, it is suggested to establish the corresponding printing settings in the printing setting data.

10. A program product, characterized in that, A computer-executed program that includes prompts for printing settings of a printing apparatus that uses cloth as the printing medium. The program enables the computer to perform the following functions: For each of the multiple fabric images obtained by photographing multiple fabrics, registration data is prepared. In the registration data, the vector representing the surface characteristics of the fabric and determined based on the fabric image is embedded to establish a correspondence with the type of fabric represented by the fabric image. For each of the multiple fabric types, corresponding printing setting data is prepared to be established between the fabric type and the printing settings of the printing device; By comparing the embedding representation of the object fabric image obtained based on the photographed object fabric (i.e., the object embedding representation) with the embedding representation contained in the registration data (i.e., the registration embedding representation), the fabric type that is the same as or similar to the object fabric is determined; and Regarding the same or similar fabric types determined for the target fabric, it is suggested to establish the corresponding printing settings in the printing setting data.

Images