JP6592755B1 - Information processing system, information processing apparatus, server apparatus, program, or method - Google Patents

Abstract

A system, method, and program for more appropriately using information related to fashion are provided. An information processing apparatus of the system includes machine learning apparatuses 22a and 22b that perform machine learning, and a terminal 23 that displays a program execution result and an information processing result. The system uses a machine learning unit that machine-learns the relationship between an image and attribute information using an image for machine learning and a plurality of attribute information included in the image for machine learning. The generated attribute information is acquired, and the element of the first database is generated by associating the acquired attribute information with the information related to the fashion image. [Selection] Figure 2

Description

  The technology disclosed in the present application relates to an information processing system, an information processing apparatus, a server apparatus, a program, or a method.

  With the recent advanced information society, there is an increasing need to properly handle fashion information on computers.

JP 2012-14544 A JP 2010-262425 A

However, in the above-mentioned patent documents, there is no system that can appropriately support using information on fashion.
Therefore, various embodiments of the present invention provide an information processing system, an information processing apparatus, a server apparatus, a program, or a method in order to solve the above-described problems.

The system of the first embodiment is
An image for machine learning and a plurality of attribute information included in the image for machine learning are used, and a neural network unit that machine-learns the relationship between the image and the attribute information by deep learning,
An acquisition unit for acquiring a first target image;
A generator that generates a plurality of attribute information related to the first target image by applying the first target image to the neural network;
A system comprising:

The system of the second embodiment is
The first target image is a reference image to be referred to;
An association unit that generates a first database by associating the information related to the reference image and the attribute information related to the reference image generated by the generation unit;
The system of a 1st embodiment.

  Here, as will be described later, the reference image has one or more advanced fashion images, popular fashion images, periodic fashion images, brand images, specific age images, specific area images, interest images, contact images, It may be a specific group-related image or an image of another specific viewpoint.

  The first database may be an attribute database corresponding to the reference image. According to the above reference image, advanced fashion image attribute database, popular fashion image attribute database, periodic fashion image attribute database, brand image attribute database, specific age image attribute database, specific area image attribute database, interest image attribute database, contact image It may be an attribute database, a specific group-related image attribute database, or the like.

  Further, the first database may not include information relating to the reference image. In this case, the first database may be generated by associating a plurality of attribute information related to the reference image generated by the generation unit.

The system of the third embodiment is
The information relating to the reference image is information relating to a fashion medium relating to the reference image.
The system of 1st or 2nd embodiment.

The system of the fourth embodiment is
A statistical processing unit for performing statistical processing using information in the first database;
A system according to any one of the systems of the first to third embodiments.

Further, the system of the fourth embodiment may not have a part related to machine learning. That is, the system of the other fourth embodiment is
A first attribute database having a plurality of attribute information related to one reference image in association with each other;
A statistical processing unit that performs statistical processing using information in the first database;
It may be a system of an embodiment provided with. In the system of the fourth embodiment, instead of having the first attribute database, an acquisition unit that acquires a part of the first attribute database from the network may be provided.

  Further, in the system of the fourth embodiment that does not have a portion related to machine learning, the first attribute database further includes information related to the one reference image associated with a plurality of attribute information related to the one reference image. May be.

The system of the fifth embodiment is
A plurality of image databases including a plurality of image data including the first target image;
A selection unit for selecting an image database including the first target image from the plurality of image databases;
A system according to any one of the systems of the first to fourth embodiments.

The system of the sixth embodiment is
An acquisition unit for acquiring a second target image;
A generator that generates a plurality of attribute information related to the second target image by applying the second target image to the neural network;
In the first database, using attribute information, a specifying unit that specifies a second fashion item that is the same as or similar to the second target image,
A system according to any one of the first to fifth embodiments.

The system of the seventh embodiment is
An acquisition unit for acquiring a third target image including selectable fashion items;
Applying the third target image to the neural network to generate a plurality of attribute information related to the third target image;
An associating unit for associating and storing the information relating to the selectable fashion item and the plurality of attribute information relating to the generated third target image in a second database;
A specifying unit that specifies, in the second database, a third fashion item that is the same as or similar to the second fashion item;
The system according to any one of the first to sixth embodiments.

The system of the eighth embodiment is
An acquisition unit for acquiring a fourth target image including selectable fashion items;
Applying the fourth target image to the neural network to generate a plurality of attribute information related to the fourth target image;
An associating unit for associating the selectable fashion item with a plurality of attribute information relating to the generated fourth target image and storing them in a second database;
A specifying unit for specifying, in the first database, a fourth fashion item different from the second fashion item in the image relating to the second fashion item;
A specifying unit that specifies, in the second database, a fifth fashion item that is the same as or similar to the fourth fashion item;
A system according to any one of the systems of the first to seventh embodiments.

The system of the ninth embodiment
An acquisition unit for acquiring a third database having a plurality of images including fashion items relating to the user;
A plurality of attribute information related to each image by applying each image related to the third database to the neural network;
An association unit for associating and storing each image and a plurality of attribute information relating to each generated image as a fourth database;
An acquisition unit for acquiring information indicating the first user image from within the third database;
A specifying unit for specifying, in the first database, a sixth fashion item that is the same as or similar to the one fashion item related to the first user image in the fourth database;
A specifying unit for specifying a seventh fashion item different from the sixth fashion item in the image relating to the sixth fashion item;
A specifying unit for specifying an eighth fashion item identical or similar to the seventh fashion item in the fourth database;
The system according to any one of the first to eighth embodiments.

The system of the tenth embodiment
The first target image is an image including a fashion item according to a first group,
The system further comprises:
A generating unit that generates a fifth database including a plurality of pieces of attribute information related to the first target image generated by the generating unit;
By applying an image including a fashion item related to a second group other than the first group to the neural network, a sixth item composed of a plurality of attribute information related to the image including the fashion item related to the second group A generation unit for generating a database;
A statistical processing unit for statistically processing the information in the fifth database and the information in the sixth database;
The system according to any one of the first to ninth embodiments.

The method of the eleventh embodiment is
Computer
Using a machine learning image and a plurality of attribute information included in the machine learning image, a machine learning step of performing machine learning by deep learning of a relationship between the image and the attribute information;
Obtaining a first target image;
Applying the first target image to the neural network to generate a plurality of attribute information related to the first target image;
Including methods.

The method of the twelfth embodiment
The first target image is a reference image to be referred to;
Generating a first database by associating the information related to the reference image and a plurality of attribute information related to the reference image generated by the generation unit;
Statistically processing information in the first database;
A method according to the eleventh embodiment, comprising:

  In addition, the method of the twelfth embodiment may not include the steps related to machine learning. That is, the method of the other twelfth embodiment may be a step of performing statistical processing using information in the first attribute database having a plurality of attribute information related to one reference image in association with each other. Further, the method of the other twelfth embodiment includes a step of acquiring a part of information in a first attribute database having a plurality of attribute information related to one reference image in association with one of the pieces of information. The method may include a step of performing statistical processing using a part or the whole.

The method of the thirteenth embodiment is
Obtaining a second target image;
Applying the second target image to the neural network to generate a plurality of attribute information of the first fashion item according to the second target image;
Identifying a second fashion item that is the same as or similar to the first fashion item according to the second target image in the first database;
A method according to the eleventh or twelfth embodiment.

The method of the fourteenth embodiment
Obtaining a fourth target image including selectable fashion items;
Applying the fourth target image to the neural network to generate a plurality of attribute information related to the fourth target image;
Associating and storing the selectable fashion item and the generated information in a second database;
Identifying in the first database a fourth fashion item different from the second fashion item in the image relating to the second fashion item;
Identifying a fifth fashion item in the second database that is the same as or similar to the fourth fashion item;
A method according to any one of the eleventh to thirteenth embodiments.

  The program according to the fifteenth embodiment is a program for causing a computer to function as the system according to any one of the first to tenth embodiments.

  The attribute information in the system, method, or program described above includes a plurality of fashion item attributes related to one fashion item, a plurality of fashion item attributes related to each fashion item in the plurality of fashion items, and a plurality of people related to one person. Attributes, a plurality of person attributes related to each person in a plurality of persons, or a plurality of person attributes related to each person in one or a plurality of persons and a plurality of fashion items in one or a plurality of fashion items related to each person It may be a fashion item attribute.

  According to an embodiment of the present invention, information on fashion can be utilized more appropriately.

FIG. 1 is a block diagram illustrating a configuration of one information processing apparatus of a system according to an embodiment. FIG. 2 is a block diagram illustrating a configuration of another information processing apparatus of the system according to the embodiment. FIG. 3 is a block diagram illustrating functions of a system according to an embodiment. FIG. 4 is an example of a system flow according to an embodiment. FIG. 5 is an example of a data structure of the system according to the embodiment. FIG. 6 is an example of a data structure of the system according to the embodiment. FIG. 7 is an example of a data structure of the system according to the embodiment. FIG. 8 is an example of a data structure of the system according to the embodiment. FIG. 9 is an example of a data structure of the system according to the embodiment. FIG. 10 is an example of a data structure of the system according to the embodiment. FIG. 11 is an example of a system flow according to an embodiment. FIG. 12 is an example of a data structure of a system according to an embodiment. FIG. 13 is an example of a data structure of a system according to an embodiment. FIG. 14 is an example of a system flow according to an embodiment. FIG. 15 is an example of a data structure of a system according to an embodiment. FIG. 16 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 17 is an example of a data structure of a system according to an embodiment. FIG. 18 is an example of a display screen of the system according to an embodiment. FIG. 19 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 20 is an example of a display screen of the system according to an embodiment. FIG. 21 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 22 is an example of a displayable table of the system according to an embodiment. FIG. 23 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 24 is an example of a graph that can be displayed by the system according to an embodiment. FIG. 25 is an example of a system flow according to an embodiment. FIG. 26 is an example of a system flow according to an embodiment. FIG. 27 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 28 is an example of a system flow according to an embodiment. FIG. 29 is an example of a system flow according to an embodiment. FIG. 30 is an example of a system flow according to an embodiment. FIG. 31 is an example of a system flow according to an embodiment. FIG. 32 is an example of a system flow according to an embodiment. FIG. 33 is an example of a system flow according to an embodiment. FIG. 34 is a block diagram illustrating a configuration of another information processing apparatus of the system according to an embodiment. FIG. 35 is a block diagram illustrating a configuration of another information processing apparatus of the system according to an embodiment. FIG. 36 is an example of a system flow according to an embodiment. FIG. 37 is an example of a display screen of the system according to an embodiment. FIG. 38 is an example of a display screen of the system according to an embodiment. FIG. 39 is an example of a display screen of the system according to an embodiment. FIG. 40 is an example of a system flow according to an embodiment. FIG. 41 is an example of a system flow according to an embodiment. FIG. 42 is an example of the data structure of the system according to an embodiment. FIG. 43 is an example of the data structure of the system according to an embodiment. FIG. 44 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 45 is an example of a graph that can be displayed by the system according to the embodiment. FIG. 46 is a block diagram illustrating a configuration of another information processing apparatus of the system according to an embodiment. FIG. 47 is an example of a system flow according to an embodiment. FIG. 48 is a block diagram illustrating functions of a system according to an embodiment. FIG. 49 is a block diagram illustrating functions of a system according to an embodiment. FIG. 50 is an example of a system flow according to an embodiment.

1. Each configuration of the information processing apparatus 10 The system of the present embodiment may be configured as shown in FIG. 1, for example. The system according to this embodiment includes an information processing apparatus 10. As an example of configuring the information processing apparatus 10, for example, as illustrated in FIG. 1, a configuration including a bus 11, an arithmetic device 12, a storage device 13, an input device 14, a display device 15, and a communication IF 16 may be used. The information processing apparatus 10 may be configured to be directly or indirectly connected to the network 19.

  The bus 11 has a function of transmitting information among the arithmetic device 12, the storage device 13, the input device 14, the display device 15, and the communication IF 16.

  The arithmetic device 12 is a device capable of executing program instructions. An example is a processor. This may be a CPU or an MPU. In addition, a graphics processing unit, a digital signal processor, and the like may be included.

  The storage device 13 is a device that records information. The storage may be temporary storage or long-term storage. The storage device 13 may be either an internal memory or an external memory, or both. The storage device 13 may be a magnetic disk (hard disk), an optical disk, a magnetic tape, a semiconductor memory, or the like. Further, the storage device 13 is not shown, but may be a storage device via a network or a storage device on a cloud via a network.

  Note that registers, L1 caches, L2 caches, etc. that store information at positions close to the arithmetic unit may be included in the arithmetic unit 12 in the block diagram of FIG. These may be included in the storage device 13. In short, it is only necessary that the arithmetic device 12, the storage device 13, and the bus 11 are configured so as to execute information processing in cooperation.

  The storage device 13 can include a program for executing a technique related to the present invention, and can also appropriately record data necessary when related to the present invention. For example, when learning data necessary for machine learning, which will be described later, is acquired, it is stored, a program for executing machine learning is stored, and attribute information generation function generated by learning after executing machine learning Information necessary for the storage may be stored. The storage device 13 may include a database. The storage device 13 may be a short-term storage device or a long-term storage device. Examples of short-term storage devices include caches, and examples of long-term storage devices include HDDs and USBs. Information generated by an attribute information generation function described later may be stored in a short-term storage device or a long-term storage device.

  Moreover, although the above demonstrated as an example the case where the arithmetic unit 12 was performed based on the program with which the memory | storage device 13 was equipped, the said bus | bath 11, the arithmetic device 12, and the memory | storage device 13 were combined. As one example, the information processing according to the system of the present application may be realized by a programmable logic device that can change the hardware circuit itself or a dedicated circuit that determines the information processing to be executed. When realized by a programmable logic device or a dedicated circuit, there is an advantage that the system of the present application can be processed at a higher speed. Further, the arithmetic device 12 may be configured to be capable of parallel calculation. For example, when the arithmetic device 12 performs machine learning, machine learning may be realized simultaneously for a plurality of learning data. Further, when the attribute information generation function is realized by the arithmetic device 12 by machine learning as described later, the attribute information generation function may be realized by parallel calculation for a reference image described later.

  The input device 14 inputs information. Examples of the input device 14 include an input device such as a keyboard, a mouse, a touch panel, or a pen-type pointing device. Note that the input device 14 may have other functions so that the touch panel has a display function.

  As an input user of the input device 14, for example, a user who performs machine learning, a user who generates information by applying a machine-learned attribute information generation function to a reference image described later, and attribute information described later is used. Examples include a user to be compared or a user who specifies a fashion item.

  The display device 15 has a function of displaying information. For example, a liquid crystal display, a plasma display, an organic EL display, and the like can be mentioned. Moreover, you may equip some input devices 14 like a touch panel.

  The network 19 has a function of transmitting information together with the communication IF 16. For example, the information processing apparatus 10 has a function of transmitting information held by the information processing apparatus 10 to another information processing apparatus via the network 19. The communication IF 16 may have any connection format. For example, USB, IEEE1394, Ethernet (registered trademark), PCI, SCSI, or the like may be used. The network 19 may be either wired or wireless. In the case of wired, an optical fiber, a coaxial cable, or the like may be used.

  The information processing apparatus 10 may be a dedicated information processing apparatus for part or all of the technology described in the present application document, or may be an information processing apparatus capable of executing a technique other than the technique described in the present application document. Also good. The information processing apparatus 10 may be a server, a workstation, a desktop personal computer, a laptop personal computer, a notebook personal computer, a PDA, a mobile phone, a smartphone, or the like.

  In FIG. 1, the information processing apparatus 10 has been described as a single information processing apparatus 10, but the information processing apparatus 10 may include a plurality of information processing apparatuses. The plurality of information processing apparatuses may be connected internally or externally via a network. Further, when the information processing apparatus 10 is composed of a plurality of information processing apparatuses, even if the owner or the administrator is different, the information processing apparatus 10 has information such as an access right when executing the technique described in this processing. What is necessary is just to be a state which can utilize an apparatus. Further, the information processing apparatus 10 may be physically present or may be virtual. For example, the information processing apparatus 10 may be virtually realized using cloud computing. FIG. 1 illustrates the basic configuration of the system according to the present embodiment. However, when the system is linked to other information processing apparatuses, the input apparatus 14 and the display apparatus 15 are not included. There may be.

  FIG. 2 shows an embodiment in which the information processing apparatus 10 described in FIG. 1 is used on a network. In FIG. 2, the system of the present embodiment includes a system server 21 that manages the whole, a machine learning device 22 a and a machine learning device 22 b that perform machine learning, and a terminal 23 that displays the results of program execution and information processing. Have. Note that when the machine learning device 22 is managed by another company on the cloud, the system according to the present embodiment may not include the machine learning device. Further, when the terminal 23 uses a general PC, the system according to the present embodiment may not include the terminal 23.

2. Example of System Function An example of the function of this system will be described with reference to FIG. 48 and 49 are other examples of the functions of the system. Each of these units may be configured to perform information processing using each hardware device described with reference to FIGS. 1 and 2. In addition, the system of the present embodiment does not always include all the following units, and may have a partial configuration in relation to a business form or a technical advantage to be realized.

2.1. Learning Data Part The learning data part has a function of acquiring data in which a learning image and information related to the learning image are associated with each other. The learning data unit may include a database including the associated data, or the associated data may be acquired from outside on the network.

2.2. Reference image portion The reference image portion has a function of providing a reference image described later. The reference image part itself may have a reference image, or a reference image may be acquired from a network. When the reference image is acquired from the network, it may be acquired from another information processing apparatus or may be acquired from the cloud. These clouds and other information processing apparatuses may be owned by the user of this system, managed, or accessible by contract. The reference image portion may include a database including a plurality of reference images (hereinafter, also referred to as “reference image set”). As described later, the reference image set includes, for example, a plurality of fashion images, a plurality of advanced fashion images, a plurality of specific age images, a plurality of specific group-related images, and the like. Moreover, the reference image part may have a plurality of reference image sets.

2.3. Reference Image Selection Unit When this system includes a reference image unit, the reference image selection unit has a function of selecting one or a plurality of reference image sets in the reference image unit. One or a plurality of reference image sets may be selected. Further, the reference image selection unit applies a set operation such as a sum, product, difference, and complement to the selected plurality of reference image sets, and the plurality of applied images are used as a reference image set. It may be used.

2.4. Target Image Part The target image part has a function of providing a target image. The provided target image may be held by the target image unit, or may be acquired from the network. When acquiring the target image from the network, the target image may be acquired from another information processing apparatus or may be acquired from the cloud. These clouds and other information processing apparatuses may be owned by the user of this system, managed, or accessible by contract.

2.5. Artificial intelligence unit The artificial intelligence unit has an attribute information generation function. The artificial intelligence unit may acquire a reference image or a reference image set and generate attribute information. For the reference image set, attribute information corresponding to each reference image may be generated. The artificial intelligence unit may generate attribute information with the target image as an input. The artificial intelligence unit may have a function of machine learning in order to improve an attribute information generation function.

2.6. Statistical processing unit statistical processing unit performs statistical processing. The result of the statistical processing may be generated only as information, or the generated information may be transmitted to another information processing apparatus on the network, or displayed on a display device connected to the system. Good.

3. Embodiment
3.1. Embodiment 1
In general, it is difficult to collect information on what kind of person wears what kind of fashion.
Therefore, the system according to the first embodiment provides a technique for collecting information related to the relationship between fashion items.

  This embodiment will be described with reference to FIG. First, a learning image is acquired in order to make machine-related information on fashion (step 1).

  The learning image may be acquired from various information sources. For example, it may be prepared from media in the fashion industry, collection photos, collection videos, top brand photos, top brand videos, top buddick brand photos, top buddick brand videos, select shop photos, select shop videos, general brand photos, general brands Examples include moving images, general consumer photos, and general consumer moving images.

  Examples of the medium include a fashion show, a designer's homepage, and a publication of a fashion association.

  Further, information related to the learning image may be stored in association with the learning image. The information related to the learning image may be, for example, the type of fashion item, the color of the fashion item, the pattern of the fashion item, the processing of the fashion item, the accessory of the fashion item, the length of the fashion item, and the like. In addition, the specific viewpoint of the property of these fashion items is called fashion item attribute in this application document. Here, the fashion item attributes are not limited to those described above. As will be described below, for example, in the type of fashion item, there are shirts, blouses, blousons, etc. as subordinate concepts under tops and outerwear. In this case, the detail classification may be a fashion item attribute for the fashion item type. The subordinate concept is not limited to the type of fashion item, but may be another fashion item attribute. As described later, the fact that one fashion item attribute has a lower-level concept fashion item attribute is sometimes referred to as hierarchization of fashion item attributes. As will be described later, the term fashion item attribute may indicate information used in machine learning, or may be generated by applying to a target image using a system after machine learning. It may also indicate attribute information.

  If it is a type of fashion item, for example, tops (T-shirts / cutsews, shirts, blouses, polo shirts, sweaters / knits, vests, hoodies, sweatshirts, cardigans, tank tops, camisole, bustiers, vests), outerwear (tailored jackets, Collarless jacket, denim jacket, riders jacket, blouson, down jacket, coat, trench coat, military jacket, down vest, knit coat, gown dress), pants (denim pants, cargo pants, chinos, slacks), dress (one piece, All-in-one), bags (shoulder bags, tote bags, backpacks / backpacks, Boston bags, body bags / waist pouches, handbags, clutch bags Suitcase / carry bag, basket bag), shoes (sneakers, slip-ons, sandals, pumps, boots, booties, dress shoes, ballet shoes / loafers, moccasins / deck shoes, rain shoes), hats (hat, beanie, hunting / beret) Cap, casket), and skirt.

  The color of the fashion item may be, for example, white, black, gray, brown, beige, green, blue, purple, yellow, pink, red, orange, gold, silver or the like.

  If it is a pattern of a fashion item, for example, it may be a plain pattern, a border pattern, a dot pattern, a stripe pattern, a check pattern, a floral pattern, a camouflage pattern, a leopard pattern, and the like.

  If it is processing of a fashion item, it may be, for example, no processing, lace processing, pleating processing, chiffon processing, gather processing, fur processing, corduroy processing, and the like.

  If it is an accessory of a fashion item, it may be without an accessory, with a frill, with a ribbon, with a bijou, with a one point, with a damage, with a point fur, etc., for example.

  If the length of the fashion item, for example, short sleeve, half sleeve, long sleeve, no sleeve, short length, middle length, long length, skinny tight, standard, wide loose, short length, above knee length, half length above knee length Full length, short / mini length, knee length, mimore / half length, long maxi length, etc.

  Here, specific values such as white, black, or gray in the color of the fashion item, and specific values such as plain pattern, border pattern, and dot pattern in the pattern of the fashion item are referred to as fashion item attribute values.

  FIG. 5 is an example in which an ID indicating a learning image and attribute information related to the learning image are associated and recorded in the database. An ID indicating each learning image is associated with one fashion item included therein. In addition, a plurality of fashion item attributes are associated with the one fashion item. In this manner, the learning image and the information related to the learning image may be stored in association with each other. Here, the fashion item attributes include the type of the fashion item, the color of the fashion item, the processing of the fashion item, and the like. Further, specific attribute values are stored for each learning ID. As for the types of fashion items, the category details 1 and the category details 2 and the categories are detailed as more detailed categories. Thus, the fashion item attribute may be hierarchized. Further, the fashion item attribute to be hierarchized is not limited to the type of fashion item, but may be another fashion item attribute or may be hierarchized. Since the fashion item attributes in the learning data are hierarchized, there is an advantage that the fashion item attributes generated by the attribute information generation device described later can also be hierarchized.

  The system of the present embodiment may be configured to generate attribute information by associating a plurality of fashion item attributes related to one fashion item with one learning image and performing machine learning as learning data. When the attribute information generation function generated in this way is applied to an image including a fashion item, there is an advantage that a plurality of fashion item attributes related to the one fashion item can be generated. When the plurality of fashion item attributes are stored in association with each other, information regarding the plurality of fashion item attributes can be used. Specifically, in a certain image, information about a polo shirt is generated as a fashion item type of a fashion item attribute, information about green is generated as a color of a fashion item of a fashion item attribute, and a fashion item of a fashion item attribute When the information about the border pattern is generated as the pattern, the polo shirt, the green, and the border pattern are stored in association with each other. This is advantageous in that a person's preference or impression of fashion has a great influence on a combination of a plurality of fashion item attributes related to a fashion item, so that these pieces of information are stored in association with each other. As a means for generating a plurality of attribute information related to a fashion item, there is a technical advantage in performing machine learning by associating a plurality of attribute information related to a fashion item.

  FIG. 6 shows another example in which the ID indicating the learning image and the attribute information related to the learning image are associated and recorded in the database. An ID indicating each learning image is associated with a plurality of fashion items included therein. In addition, a plurality of fashion item attributes are associated with each fashion item. In this way, the learning image and the attribute information related to the learning image may be stored in association with each other.

  The system of this embodiment may associate one fashion item with one learning image and use it as learning data. However, the system according to this embodiment provides a plurality of fashions for one learning image. The attribute information may be generated by associating items and machine learning as learning data. For example, when a top image, a skirt, and a bag are included in an image for learning, information on the fashion item names, colors, patterns, processing, accessories, etc. of these tops, a skirt, and a bag. The information may be associated with the learning image. When the attribute information generation function generated in this way is applied to a target image in which a plurality of fashion items are included in one image, a plurality of fashion item attributes related to the plurality of fashion items are displayed. There is an advantage that it can be generated. In the case where a plurality of fashion items are included in the image to be identified, the two or more fashions are generated by generating a fashion item attribute related to at least two or more fashion items of the plurality of fashion items. There is an advantage that the relationship between items can be generated in association with information. In general, since a combination of fashion is taken into consideration, there is an advantage that information can be generated from an image including a relationship between a plurality of fashion items and a relationship between a plurality of fashion item attributes related to each fashion item.

  The attribute information related to the learning image may be information related to a person included in the learning image. For example, it may be information indicating a person from a specific viewpoint, such as gender, age, age group, height, physique, and facial expression. These are called human attributes in the present application documents. The specific value is called a human attribute value. As attribute information, the human attribute has the advantage that information about fashion items can be enriched. In other words, whether fashion items are favored by men, favored by women, by what age group, and by what height It is also related to information about people about what physique they like. Therefore, there is an advantage that it is possible to generate information about what type of person is associated with a fashion item by storing such a person attribute in association with the fashion item. Moreover, there is an advantage that by storing such a person attribute in association with a fashion item attribute, it is possible to generate information as to what type of person the fashion item attribute is associated with. In addition, the person attribute may be hierarchized similarly to the above-described fashion item attribute. For example, the age group in increments of 10 years, the age group in increments of 5 years, and the age group in increments of 2 years are hierarchized in the age group, or the age group in increments of 5 years and a specific age In this way, the age group and the age may be hierarchized. In addition, as with the fashion item attribute, the human attribute may indicate information used in machine learning, or the result obtained by applying it to the target image using the system after machine learning. The corresponding information may be indicated. Here, it may be referred to as “attribute information” as including a fashion item attribute and a human attribute, and may be referred to as “attribute value” as including a fashion item attribute value and a human attribute value. The attribute information may also be hierarchized.

  FIG. 7 shows another example in which an ID indicating a learning image and attribute information related to the learning image are associated and recorded in the database. An ID indicating a learning image is associated with a person and a fashion item included in the learning image. Each person and fashion item may be one or more. People and fashion items may also be associated. The association between a person and a fashion item may be a relationship between fashion items worn by a person, but is not limited to this relationship. In addition, a plurality of person attributes are associated with each person. In addition, a plurality of fashion item attributes are associated with each fashion item. In this manner, the learning image and the information related to the learning image may be stored in association with each other.

  FIG. 8 shows still another example in which an ID indicating a learning image and attribute information related to the learning image are associated and recorded in the database. It is an example of a relational database, and is associated through a learning image ID and a person in the learning image. For example, for the learning image ID 1002, there are three people in the image, and one or more fashion items are associated with each person, and a plurality of fashion item attributes are associated with each one fashion item. It has been. In the case of FIG. 8, for example, the number of records including one attribute value of a fashion item type indicates the number of fashion items, and one reference image includes one fashion item. There is an advantage that it is possible to handle a case where a plurality of fashion items are included in a unified manner. In addition, in the human attribute of FIG. 8, an example in which attribute information of a rough age group and a fine age group is provided and the human attribute is hierarchized is shown.

  In the above description, as a relationship between a person and a fashion item, the fashion item captured in the image is associated with the position worn by the person and the person, but the relationship between another person and the fashion item. It may be attached. For example, a mode in which a person holds a fashion item by hand, or a person is located behind the fashion item. In this way, if a person can find a relationship between a person and a fashion item when he / she sees the image, it is possible that a rule related to the relationship that a person can find can be derived by machine learning. A person and a fashion item may be associated based on a machine-learned rule.

  Next, the system of the present embodiment may perform machine learning to generate an attribute information generation function (Step 2). As input information, information in which a learning ID based on the ID indicating the learning image described above and attribute information related to the learning image are associated may be used. Thus, the relationship between the learning image and the attribute information related to the learning image may be machine-learned. The system of the present embodiment may have an attribute information generation function capable of generating attribute information related to an image input by machine learning.

  Various artificial intelligence techniques may be used for machine learning. Examples of artificial intelligence technologies include machine learning technologies such as neural networks, genetic programming, functional logic programming, support vector machines, clustering, regression, classification, Bayesian networks, reinforcement learning, expression learning, decision trees, and k-means. May be used. In the following, an example using a neural network is used, but it is not necessarily limited to a neural network.

  As the machine learning technique using the neural network, a deep learning technique may be used. This is a technique that enables output even for unknown inputs by learning the relationship between inputs and outputs using a plurality of layers. There are supervised and unsupervised methods, either of which may be applied. In FIG. 7 described above, the ID indicating the learning image is associated with the attribute information related to the learning image, and the information related to the learning image is given as the teacher. Although it is an example with existence, it is not necessary to use the method with a teacher, and the method without a teacher may be used.

  The system according to the present embodiment may perform machine learning on learning data in which a learning image and a plurality of fashion item attributes related to one fashion item included in the learning image are associated with each other. Machine learning may be performed on learning data in which an image for use and a plurality of fashion item attributes related to a plurality of fashion items included in the learning image are associated with each other.

  Further, the fashion item attribute and the human attribute for the learning image may be collectively machine-learned in one machine learning device, or the machine learning for the fashion item attribute related to the learning image and the human attribute related to the learning image The machine learning may be performed separately. When machine learning about the fashion item attribute related to the learning image and machine learning about the human attribute related to the learning image are performed separately, it is possible to focus on each attribute and perform more efficiently. Machine learning is possible.

  Moreover, the apparatus which carries out machine learning only about a fashion item attribute may be sufficient. This is because, among the subjects learned by the human attribute, the age and sex that are the human attributes may be collected from external information of the image.

  The system of this embodiment after machine learning may be configured to output attribute information related to the target image when applied to the acquired target image. The attribute information may be one or a plurality of fashion item attributes related to one or a plurality of fashion items in the target image, or one or a plurality of person attributes related to one or a plurality of people in the target image. It may be one or a plurality of fashion item attributes related to one or a plurality of fashion items in the target image and one or a plurality of person attributes related to one or a plurality of people in the target image. FIG. 9 is an example of the generated attribute information. Here, a human attribute and a fashion item attribute are associated as attribute information. There may be one or more fashion items. Further, attribute information generated by the attribute information generation function may be hierarchized by hierarchizing the attribute information of the learning database. By layering attribute information, more detailed attribute information can be generated for the target image.

  In addition, when the system of the present embodiment after machine learning is applied to a target image including a plurality of people, one or a plurality of attributes relating to one or a plurality of fashion items in the target image, and / or A plurality of attributes relating to a plurality of persons in the target image may be output. Here, the attribute which concerns on one or several fashion items may be linked | related with the person who wears the said fashion item, and may be linked | related with the said person's attribute. FIG. 10 relates one or a plurality of person attributes and one or a plurality of fashion item attributes related to one or a plurality of fashion items to each of a plurality of persons in the target image based on the order described. This is an example generated.

  Further, by combining the above, when the system of the present embodiment after machine learning is applied to the acquired target image, one or more attributes relating to one or more fashion items in the target image, and Alternatively, a plurality of attributes relating to one or more persons in the target image may be output.

  The generated information may be stored as appropriate.

The system of this embodiment is
An acquisition unit for acquiring an image for machine learning including one or more fashion items in the image;
A neural network unit that machine-learns the relationship between the one learning image and one or more fashion item attributes related to the one or more fashion items by deep learning;
The attribute information generation system may be provided.

  When the system of this embodiment is provided with the above-described configuration, there is an advantage that one or more fashion item attributes that are information on one or more fashion items can be acquired from an image including one or more fashion items.

The system of this embodiment is
An acquisition unit for acquiring a machine learning image including a person and one or more fashion items in the image;
A neural network unit for machine learning by deep learning of the relationship between the one learning image, the human attribute relating to the person, and the one or more fashion item attributes relating to the one or more fashion items;
The attribute information generation system may be provided.

  When the system of the present embodiment has the above-described configuration, one or more fashion items that are information on one or more person attributes and one or more fashion items from an image including a person and one or more fashion items. There is an advantage that attributes can be obtained in association with each other.

The system of this embodiment is
An acquisition unit for acquiring an image for machine learning including one or more people and one or more fashion items in the image;
The one learning image, one or more person attributes relating to the one or more persons, and one or more fashion item attributes relating to the one or more fashion items associated with the one or more persons And the neural network part that performs machine learning by deep learning, and
The attribute information generation system may be provided.

  In addition, the network unit in the document of the present application is the machine learning image including one or more people and one or more fashion items, and the one learning image and one or more people related to the one or more people. Machine learning of the relationship between the attribute and one or more fashion item attributes related to the one or more fashion items associated with the one or more people by deep learning to produce a learned neural network unit (Including the method of manufacturing the neural network unit, it is expressed in the present application as production). The neural network unit may be a program.

  When the system of the present embodiment has the above-described configuration, one or a plurality of attributes relating to one or a plurality of people, and each of the plurality of people from an image including one or a plurality of people and one or a plurality of fashion items. There is an advantage that one or a plurality of person attributes associated with a person and one or a plurality of fashion item attributes that are information about the one or more fashion items associated with each person of the person can be acquired.

  As described above, the system of the present embodiment includes one or more people and / or one or more fashion items in one learning image, and is associated with the one learning image. The machine learning device that has learned one or a plurality of person attributes related to the one or more persons and / or one or more fashion item attributes related to the one or more fashion items as learning data is attribute information. Can be generated. Since the system of this embodiment having the attribute information generation function can generate attribute information when applied to an image including a fashion item, information on the fashion item is organized from the image including the fashion item. There is an advantage that can be acquired.

3.2. Embodiment 2
It may be beneficial for fashion industry officials or general consumers to refer to fashion information from professionals and consumers. However, these pieces of information are not necessarily organized, and it may be difficult to use the information. Therefore, the system according to the present embodiment provides a technique for organizing information on an image (hereinafter referred to as “reference image”) relating to such information to be referred to (hereinafter referred to as “reference information”). .

  The hardware configuring the system of the present embodiment may be the same as the hardware described in the first embodiment. Note that the information generated by the portion related to the statistical processing may be stored in the storage device 13. Further, the information may be displayed on the display device 15 or may be transmitted to the network 19 via the communication IF 16. The transmitted information may be information assumed to be displayed by another information processing apparatus.

  The system of this embodiment will be described with reference to FIG. First, a reference image set is acquired (step 1). Here, as described above, the reference image set, that is, the plurality of reference images may be sequentially acquired on the network.

  Next, the attribute information generation function of the system of the present embodiment is used for the reference image set (step 2). Here, regarding the reference image, the information generated by the attribute information generation function is referred to as “attribute information related to the reference image”. The attribute information may be generated by a machine learned device. For example, the attribute information may be generated by the function described in the first embodiment.

  Note that the attribute information generation function may be individually applied to each reference image, or the attribute information generation function may be applied to a plurality of reference images at the same time by parallel calculation or the like.

  Next, the system of the present embodiment associates and stores information related to each reference image and attribute information related to each reference image (step 3). In this step 3, the information related to the reference image and the attribute information related to the reference image are referred to as “reference image attribute database”. Since the reference image may include various images, the reference image attribute database may be simply referred to as “attribute database”.

  FIG. 12 is an example of the reference image attribute database. In FIG. 12, attribute information is stored in association with the reference image ID. In addition, as attribute information, a person attribute and a fashion item attribute are associated. In addition, attribute information generated by the machine-learned attribute information generation function may be hierarchized by attribute information of the learning database being hierarchized. As attribute information is hierarchized, more detailed attribute information can be generated.

  Note that the system according to the present embodiment may include information on each reference image. FIG. 13 is an example of a database of information related to the reference image. The information related to the reference image may be information such as an ID indicating the reference image, information about the source of the reference image, information about a person in the reference image, information about the creation time of the reference image, and the like. The information related to the source of the reference image may be the medium name of the medium on which the reference image first appears, the creation company of the medium, or the author of the medium. The information about the person in the reference image may be the person's age, age group, age, gender, and the like.

  In the reference image attribute database, the information related to the reference image may be only the ID indicating the reference image as described above, or may include information related to other reference images. Among the reference image attribute databases, a database including information related to a reference image other than an ID indicating a reference image is particularly referred to as an “enlarged reference image attribute database”. Note that the enlarged reference image attribute database includes a relational database system that links the attribute information database and the information database related to the reference image other than the ID indicating the reference image through the reference image ID. And various database techniques may be applied.

  In the above description, the configuration in which the attribute information and the information related to the reference image are stored in association with each other has been described. However, the configuration is not limited to the example. For example, in the statistical processing described later, the attribute information is simply aggregated, and if the information related to the reference image is not used, the information related to the reference image may not be associated.

  A user who wants to refer to the reference image can perform various analyzes on the reference image by using the reference image attribute database. For example, by performing statistical processing on the information in the reference image attribute database, information related to the reference image can be acquired more.

  In the statistical processing, some or all of the information in the reference image attribute database may be displayed as numerical values, or may be processed and displayed. By displaying a part, unnecessary information in relation to the display can be deleted, and the viewer's understanding can be promoted. In addition, as a processing mode, a certain calculation may be applied to information or may be graphed. Certain operations may include averaging and subtraction at the same time. For graphing, an appropriate method may be applied in relation to the display purpose.

<Application to fashion images>
There may be various reference images. For example, fashion images of collections, top brands, overseas celebrities, etc. in the fashion industry, or fashion images of reader models, domestic shop clerk, general consumers, etc. may be used. These fashion-related images (hereinafter referred to as “fashion images”) may be used as a reference when viewing the overall trend of fashion. When the system of this embodiment uses a fashion image as a reference image, it can generate fashion trend information included in the fashion image.

  Examples of fashion images include collection photos, collection videos, top brand photos, top brand videos, top buddick brand photos, top buddick brand videos, select shop photos, select shop videos, general brand photos, general brand videos, general consumers Images such as photographs and general consumer videos. Examples of media for these materials include magazines, videos, interviews, security cameras installed in stores, and in-store photography cameras. Note that a plurality of fashion images (hereinafter also referred to as “fashion image set”) may or may not include all the types of images described above. Any image regarding these fashions may be used.

  The system of the present embodiment may have a fashion image set in advance, or may acquire a fashion image set via another network. A medium from which fashion images are based may be purchased from a company that handles fashion images, or images on a network may be collected by crawling technology or the like. For example, a collection target may be a home page or an SNS (Social Network Service). There are various types of SNS, but they may be, for example, Instagram or Twitter.

  The fashion image set may be collected from the same type of medium as the learning image, or may be collected from a different type of medium. A part of the fashion image set may be the same as a part of the learning image, but is preferably different.

  The flow of the system of this embodiment when a fashion image set is applied is as shown in FIG.

  When a fashion image set is used, in step 3, the system of the present embodiment associates information indicating each fashion image with information related to each fashion image and stores it in the fashion image attribute database (step 3). .

  FIG. 15 is an example of a database that stores information indicating fashion images and information related to the fashion images in association with each other. The information related to the fashion image includes, for example, an ID indicating the fashion image, the fashion image itself, the time when the fashion image was published, the name of the brand that is the source of the fashion image, the type of brand that is the basis of the fashion image, and Or it may be the gender, age information, etc. of the person in the fashion image. Information regarding these fashion images may be given in advance.

  The user can analyze the brand by using the fashion image attribute database. Specifically, the system of the present embodiment may perform statistical processing using the fashion image attribute database (step 4). As described above, the statistical processing may display numerical values using some or all of the information in the fashion image attribute database, or may process and display them.

  For example, FIG. 16 is an example of a screen displayed after statistical processing. The vertical axis represents the number of times the corresponding fashion item is included in the reference image. For example, it is the number of reference images including tops, outerwear, pants, and the like, which are specific contents within the type of fashion item. The number of times may be the number of fashion items, the number of reference images, or a number calculated by a method of counting those having a predetermined relationship among them as one, and the calculation method is, for example, within one reference image. If one fashion item is included, the number may be the number of reference images. If one reference image includes multiple fashion items, the number is the number corresponding to each fashion item. Good. From the viewpoint of facilitating the calculation of the number of fashion items when a plurality of fashion items are included in one reference image, for example, as shown in FIG. One fashion item may be input. The horizontal axis is the time axis, and is the axis for each year, spring, summer, autumn, and winter. In other words, on the horizontal axis, the reference images that were announced during the period from March to May as spring, from June to August as summer, from September to November as autumn, and from December to January as winter are targeted. It is calculated as. The numerical value may be the total number of these periods or an average value.

  Moreover, it may replace with the classification of the fashion item in this example, and may show another fashion item attribute. For example, as described above, the color of the fashion item, the pattern of the fashion item, the processing of the fashion item, the accessory of the fashion item, the length of the fashion item, and the like may be used.

  The statistical processing unit may generate a numerical value obtained by statistically processing specific attribute information for a specific period using an attribute database. For example, as described above, the statistical processing unit may generate a summed numerical value for each specific fashion item attribute such as tops, outer, and pants of the type of fashion item. In addition, you may generate a total value for each T-shirt, shirt, blouse, etc. in the tops of the type of fashion item, plain pattern or border pattern of the pattern of the fashion item, green or black of the color of the fashion item A total number may be generated for each of the above. Further, the object to be summed may be the number of fashion items or the number of reference images including fashion items. In addition, the sum may generate a numerical value that is summed up for each predetermined period or each predetermined period such as each period of spring, summer, autumn, and winter, or an average value for each predetermined period or every predetermined period. It may be generated. The predetermined period may be the most recent one month, the most recent three months, a specific one month, a specific three months, or the like. These generated numerical values may be displayed. The attribute information described above may be a fashion item attribute or a human attribute as described above.

  Further, the statistical processing unit may be configured to present FIG. 18 and allow the viewer to select one or more attribute information to be displayed in time series.

  That is, the statistical processing unit may generate a total value for each selected attribute information. Similarly to the above, the total may be a predetermined period such as each period of spring / summer / autumn / winter or each month, or a total number may be generated every predetermined period, or the predetermined period or the predetermined period may be generated. An average value may be generated every time.

  Further, the vertical axis may be generated or displayed instead of the numerical value related to the reference image. For example, in the case where the type of the fashion item is the target and the whole is 100%, the proportion of each element in the type of the fashion item may be indicated along the time axis as described above. Here, combinations within the category of fashion items may be combined as a realistic possibility. For example, pants and outerwear, pants and tops, pants and outerwear and tops, skirts and outerwear, skirts and tops, skirts and tops and outerwear, one piece, one piece and outer, one piece and tops, outer, and the like. Here, the others are images including only a part such as the upper body and the lower body of the body. FIG. 19 is an example of such a combination. In this way, by displaying the overall proportion, it is possible to display the proportion of each of the fashion item types relative to the whole, and the viewer can display the overall proportion of the fashion item types. Understand the relative proportion of

  The statistical processing unit may generate a ratio for each attribute information when all target attribute information is 100%. The generated percentage may be displayed. The predetermined period or every predetermined period may be the same as described above.

  In addition, as a display mode, a two-axis graph of one attribute information and time can be used to display a transition according to time for the one attribute information. It is possible to easily understand the temporal transition in terms of attribute information.

  In addition, the statistical processing unit may set the reference image target on which a numerical value is generated or displayed in the drawing as data having one specific attribute information among the data included in the attribute database. For example, in a fashion image set included in the fashion image attribute database, only a fashion item having shoe attribute information as a fashion item type, or a fashion item having skirt attribute information as a fashion item type Items may be targeted. In this way, by limiting the types of fashion items such as shoes only, skirts only, temporal transitions in specific fashion item types can be displayed. Thereby, for example, when the viewer is in a position to consider a design or sales strategy related to shoes or skirts only, the fashion item can be displayed by displaying the type of a particular fashion item such as shoes or skirts only. Design and sales strategies that are limited to these types can be planned.

  In addition, the statistical processing unit may set the reference image target on which a numerical value is generated or displayed in the drawing as data having a plurality of specific attribute information in one attribute information among data included in the attribute database. For example, in one attribute information called a type of fashion item, a configuration may be adopted in which fashion items that are a plurality of attribute information such as pants and skirts can be selected. With this configuration, it is possible to display the types of fashion items having a certain commonality. Similarly, in one attribute information of the type of fashion item, items other than tops or bottoms such as bags, shoes, and hats may be targeted, and conversely tops and bottoms may be targeted.

  As described above, when the type of a specific fashion item is used as part of the data and the attribute information to be displayed is the color of the fashion item, the color transition in the type of the specific fashion item is displayed. You can also

  In addition, in the case where the object of the fashion image to be displayed in the figure is a part of the data that the fashion image attribute database has, instead of the type of the fashion item, another one or more fashion item attributes are used as a part of the data. May be shown. For example, as described above, the color of the fashion item, the pattern of the fashion item, the processing of the fashion item, the accessory of the fashion item, the length of the fashion item, and the like may be used.

  In addition, the statistical processing unit may select one or a plurality of specific attribute information in each of the attribute information in a plurality of attribute information among the data of the attribute database as a reference image target for which a numerical value is generated or displayed in the figure. It is good also as data which have. For example, numerical values may be generated by summing up tops and bottoms in one attribute information of fashion item type and black and white in other attribute information of fashion item color. The predetermined period or every predetermined period may be the same as described above. Moreover, the generated numerical value may be displayed. As described above, by limiting the targets to be generated or displayed for a plurality of attributes, the limited items can be investigated.

  The system of the present embodiment may be configured to present FIG. 20 and allow the viewer to select in order to select attribute information used for a graph to be displayed.

  That is, the statistical processing unit may generate a numerical value summed for each specific attribute information or for each selected attribute information only for data having the selected attribute information or attribute information as attribute information. Similarly to the above, the total may be a predetermined period such as each period of spring / summer / autumn / winter or each month, or a total number may be generated every predetermined period, or the predetermined period or the predetermined period may be generated. An average value may be generated every time.

  In addition, the statistical processing unit may use information related to the reference image instead of the above-described attribute information, or may mix and use both attribute information and information related to the reference image. In other words, the statistical processing unit may limit the target for one piece of information related to the reference image, may sum the numerical values, or may display the generated numerical values.

  Here, when the attribute information and the information related to the reference image handle the same type of information, the statistical processing unit may use either one or both. As an aspect using one, it may be determined in advance as attribute information or information relating to a reference image, or may be determined based on other information. The other information may be, for example, information such that one information has higher reliability than the other information. Further, as an aspect in which both the attribute information and the information related to the reference image are used, an average value of the attribute information and the information related to the reference image may be used.

  As a specific example in which the attribute information and the information related to the reference image handle the same type of information, for example, information on a person's age group can be cited. In this case, when there is information on the person's age group in the attribute information, particularly in the person information, and information on the person's age based on the information related to the reference image, the information on the person's age group in the attribute information is used. Alternatively, information on the age group of the person in the information related to the reference image may be used. At this time, for example, it may be determined that the information related to the reference image is determined from the purchase age group of the magazine and is highly reliable, while the information related to the reference image is manually input. Therefore, it may be judged that the reliability is low. In addition, the reliability information is included as one of the information related to the reference image, and if the reliability information is higher than a predetermined threshold, the information related to the reference image is determined to be high in reliability. And the information in the attribute information may be used if the reliability information is not high as compared with a predetermined threshold value.

  Another specific example in which attribute information and information related to a reference image handle the same type of information is the gender of a person. In this case, when there is information on the gender of the person in the attribute information, particularly in the human information, and information on the gender of the person based on the information related to the reference image, the information on the gender of the person in the attribute information may be used. Alternatively, information on the gender of the person in the information related to the reference image may be used. As for the gender of a person, similarly to the above, information related to reliability may be provided as one of information related to the reference image, and based on such information, attribute information and information related to the reference image may be provided. You may decide which one to use.

  In the above description, the time is assumed to be one axis of the graph. However, the present invention is not limited to the time, and a certain period may be targeted and one axis of the graph may not be the time. Here, as a fixed period, if you want to view the latest situation, you can select a period such as the past six months, for example, and add or average these values, or consider the season in which fashion is affected. Then, a specific period of spring, summer, autumn and winter may be displayed as a total or average value.

  Then, one axis of the graph may be attribute information instead of the time. For example, a two-dimensional graph may be displayed from two viewpoints. For example, FIG. 21 shows the one attribute information as the type of fashion item such as tops, outer, pants, and one piece, and the other attribute information as the color of the fashion item such as white, black, and gray. And a numerical value of the number of fashion items or the number of reference images having the same attribute information in common, and a two-dimensional graph based on these. This graph makes it possible to understand the status of each type of fashion item for each color. That is, the statistical processing unit may generate a numerical value of the number of fashion items or the number of reference images that have one attribute information and other attribute information in common, and generate a graph based on these. Similarly, instead of these two items, two items may be selected and displayed according to the browsing purpose from information relating to other fashion item attributes and brands. In addition to two dimensions, as a third viewpoint, a three-dimensional graph or higher-order graph may be used.

  Further, not limited to graphing, numerical information may be displayed as it is. For example, FIG. 22 is an example of numerical information displayed in terms of the color of a fashion item and the pattern of the fashion item. In this case, the display form of numerical information to be noted may be changed according to the browsing purpose. For example, the display form of numerical values satisfying a predetermined condition may be changed and displayed. In FIG. 22, a numerical value exceeding 400 is a tanned and bold double line. Thus, the viewer can easily pay attention to the numerical value or the target satisfying the predetermined condition by making the display mode different from the other numerical values or the target. The target for changing the display mode is not limited to the target numerical value, but the display mode of attribute information related thereto may be changed. In addition, as an example of changing the display mode, it may be a change of font, style, underline, color, size, character decoration, arrangement, etc., about the format of numerical values or characters, For the background, it may be a change in a fill or pattern.

  The statistical processing unit may generate the total number of fashion items or reference images having one attribute information and other attribute information. The statistical processing unit may display the numerical value. Moreover, a statistical processing part may display the display mode of the said numerical value in the aspect different from the numerical value which does not satisfy the said predetermined condition, when the said numerical value satisfies a predetermined condition. The different modes may be different modes such as font, style, underline, color, size, text decoration, arrangement, etc. for the format of numerical values or characters, and are filled in the columns and backgrounds related to the numerical values or attribute information. The pattern may be different.

  Information may be processed in numerical display or graph display. For example, in fashion, there is a case where it is necessary to consider every season since it varies greatly due to seasonal factors. Each season may be, for example, spring, summer, autumn and winter, or a month corresponding to these.

  In order to deal with seasonal factors, in the above numerical display or graph display, the numerical value of the same period of the previous year may be subtracted from the above numerical value, or the numerical value of the above numerical value and the same period of the previous year may be displayed. The ratio may be displayed. FIG. 23 is an example of displaying the ratio between the numerical value and the numerical value of the same period of the previous year. The same period is calculated as the ratio divided by the same month of the previous year. In FIG. 23, the number of target images is counted every two months, such as January, February, March, and April, and a line graph is created with the colors of fashion items such as white, black, and green. The vertical axis shows the ratio with the same period of the previous year. In July and August, the number of red is about 6 times and the number of green is about 5 times. By displaying as a ratio with the previous year in this way, it is possible to investigate the trend of fashion by eliminating seasonal factors. Here, the ratio is shown as a ratio with the previous year, but the same period of the previous year may be subtracted.

  The statistical processing unit may generate a numerical value obtained by subtracting the numerical value corresponding to the period corresponding to the period in the year prior to the year from the numerical value, using the year and the period used for the numerical value. . The numerical value may be displayed. The year before the year may be the previous year of the year.

  The statistical processing unit may generate a numerical value obtained by dividing the numerical value by a numerical value corresponding to the period corresponding to the period in the year before the year, using the year and the period used for the numerical value. . The numerical value may be displayed. The year before the year may be the previous year of the year.

  Further, relative numerical information may be generated for the color of the fashion item attribute. For example, in FIG. 24, 3.4%, 4.3%, 4.8%, and 5.6% indicate the relative proportion of the green color in the overall color in each year period. . When the system of the present embodiment displays such a diagram, the viewer can understand that the proportion of the green color is relatively increased in the overall color.

  The statistical processing unit may generate a numerical value of a ratio of a numerical value related to one or a plurality of attribute values in one attribute information in the one attribute information. Here, the numerical value may be displayed. In addition, a graph based on the numerical value may be displayed. The numerical value of the ratio may be a numerical value obtained by dividing the numerical value related to the one or more attribute values by the sum of the numerical values related to the attribute values included in the one attribute information.

  Further, the statistical processing unit may predict a numerical value related to the attribute value in the future using the value of the reference image attribute database. The prediction may use a numerical value for a certain period such as the latest three months or the latest six months. For example, a simple average method in which an average value of the numerical values for a certain period is set as a future numerical value may be applied, or a method for reducing seasonal factors such as a moving average method or a weighted moving average method using the numerical values for the certain period May be applied, or a method of predicting a value in a future period by generating a linear equation using the numerical value of the certain period may be applied.

  In the above description, the attribute information has been given in advance. However, the attribute information may be configured such that the category can be changed. Specifically, a plurality of attribute values may be changed to one attribute value, and a sum of numerical values related to the plurality of attribute values may be generated as a numerical value related to the changed one attribute value.

  For example, regarding the color of a fashion item, if there are attribute values of white, black, gray, brown, beige, green, blue, purple, yellow, pink, red, orange in the attribute database, Create light and dark colors, for example, change white, green, blue, yellow, pink, red, orange to light and change black, gray, brown, beige, purple to dark May be generated.

  Thus, by changing a plurality of attribute values to one attribute value, there is an advantage that global information can be easily acquired and is easily understood.

  The above-mentioned displays are materials for fashion item designers, those who adjust fashion item production, those who identify popular fashion items when selling fashion items, and those who predict fashion item trends. And so on.

  As a part, when the system of the present embodiment is applied to reference images that are considered to belong to a certain group, it is possible to extract information on the limited fashion tendency.

  The above fashion images may be grouped from any viewpoint or point of view. For example, collection / designer images, celebrity images, domestic celebrity images, domestic reader model images, domestic shop clerk images, domestic general consumers Consider a case in which images are classified.

  Here, examples of the collection / designer image include images related to fashion designers and fashion shows.

  An example of a celebrity image is an image of a celebrity who is considered to be advanced in fashion.

  A domestic celebrity image is an image relating to people who are regarded as talents or entertainers.

  The domestic reader model image is an image that is an example of a general person but wearing fashion in a fashion magazine.

  A domestic shop clerk image is an image relating to a clerk in a fashion-related shop.

  The domestic general consumer image is an image relating to a general person.

  Domestic reader model images and domestic shop clerk images may be provided by fashion manufacturers and dealers as part of sales promotions.

<Advanced fashion image>
Collection / designer images and celebrity images are collectively called “advanced fashion images”. Advanced fashion images are often related to overseas fashion, but may include domestic fashion. Advanced fashion images are often handled mainly by advanced fashion designs.

  The flow when the system of this embodiment acquires an advanced fashion image as a reference image is as shown in FIG. Here, in the system according to the present embodiment, when an advanced fashion image is used as a reference image, an “advanced fashion image attribute database” can be used instead of the fashion image attribute database when a fashion image is used. And the statistical processing part in the system of this embodiment may perform the statistical processing similar to the statistical processing part in the above-mentioned fashion image based on an advanced fashion image attribute database. The information generated by the statistical processing or the display based on the information becomes information based on advanced fashion. Therefore, the above-mentioned various tendencies can be displayed particularly in advanced fashion.

<Popular fashion image>
Domestic reader model images, domestic shop clerk images, and domestic consumer images are collectively referred to as “popular fashion images”. The popular fashion image is often handled mainly by a popular fashion design.

  The flow when the system of the present embodiment acquires a popular fashion image as a reference image is as shown in FIG. Here, in the system according to the present embodiment, when a popular fashion image is used as a reference image, a “popular fashion image attribute database” can be used instead of the fashion image attribute database when a fashion image is used. And the statistical processing part in the system of this embodiment may perform statistical processing similar to the statistical processing part in the above-mentioned fashion image based on the popular fashion image attribute database. The information generated by the statistical processing or the display based on the information is information based on popular fashion. Therefore, it is possible to display the above-described various tendencies in popular fashions such as the design for general consumers and the adjustment of the production amount related to the sales channels for general consumers.

  Note that the trend of advanced fashion images may affect popular fashion images after a certain period of time. In our survey based on limited data, there is an example that the trend of advanced fashion images affected popular fashion images after about 1.7 years. Therefore, by comparing the trend of advanced fashion images with the trend of popular fashion images, it may be a future reference or prediction of popular fashion images. FIG. 27 shows an example in which statistical information based on advanced fashion images is compared with statistical information based on popular fashion images. In this example, the white color indicated by the dotted line displays a prominent value in the summer of 2016 in statistical processing based on advanced fashion images, while in the spring of 2018 in statistical processing based on popular fashion images. The prominent value is displayed at. The black color indicated by the straight line shows a prominent value in the winter of 2016 in statistical processing based on advanced fashion images, whereas in the statistical processing based on popular fashion images, the black color prominent in autumn 2018. The value is displayed.

In addition, although not displayed in FIG. 27, with respect to corresponding information such as white or black, in the display of the advanced fashion image and the popular fashion image, a portion having a numerical value greater than a predetermined numerical value that is considered to protrude is displayed. It is good also as a structure displayed by the aspect which makes it notice that the content of the advanced fashion image affected the popular fashion image by making it display in relation. If the system of this embodiment displays the statistical information based on the advanced fashion image and the statistical information based on the popular fashion image in comparison with each other, it can be a future reference or prediction of the popular fashion image.
<Periodic fashion image>
Fashion trends in the fashion industry are said to have periodicity. Therefore, the system of the present embodiment may be mainly targeted for images that are referred to based on periodicity (hereinafter referred to as “periodic fashion images”). As an example of a “periodic fashion image”, when focusing on a cycle of 12 years, a predetermined period including 12 years ago (for example, a period from January to December 12 years ago, or 12 years) Examples include fashion images that have been published on a medium issued in the previous applicable seasonal period) or on a medium targeted for that period.

  The flow when the system of this embodiment acquires a periodic fashion image as a reference image is as shown in FIG. Here, in the system according to the present embodiment, when a periodic fashion image is used as a reference image, a “periodic fashion image attribute database” can be used instead of the fashion image attribute database when a fashion image is used. And the statistical processing part in the system of this embodiment may perform the statistical process similar to the statistical processing part in the above-mentioned fashion image based on a periodic fashion image attribute database. The information generated by the statistical processing or the display based on the information becomes information on a predetermined previous fashion from the present. Therefore, the viewer can consider the formulation of a fashion design and adjustment of the production amount relating to the sales channel of fashion items, taking into consideration the periodicity of 12 years.

  Note that the system of the present embodiment may display to make the fashion periodicity easier to understand. For example, the statistical processing unit may display the information based on the periodic fashion image and the information based on the fashion image in a period corresponding to the information based on the periodic fashion image in comparison. For example, the system according to the present embodiment is applied to a medium issued for a predetermined period 12 years ago from now, for example, a summer period from July to September, or a medium for fashion items in the period. Periodic fashion image attribute database is generated by using images including the posted images as a periodic fashion image, and two types of fashion item attributes (for example, tops and outer as the fashion item type, each color as the color of the fashion item) Suppose you want to aggregate and generate numerical values and display them. Similarly, as a period corresponding to the above, an image including an image published on a medium issued in the summer period from July to September of this year or a medium targeting fashion items in the period is targeted. An attribute database is generated, and the same fashion item types as described above are aggregated to generate and display numerical values. As a mode of display in comparison, a configuration in which the axes are the same and a mode of displaying on the same screen or a mode of displaying side by side or vertically may be used. In this way, by displaying the fashion information before the period and the current fashion information for a given period in comparison, the possibility of a change in fashion at the same time is based on the idea that fashion is periodic. The material for studying can be easily obtained. Moreover, in the above, in the numerical values calculated based on the periodic fashion image attribute database, information different from the numerical values not higher than the predetermined numerical values may be added to numerical values higher than the predetermined numerical values. Further, in the aggregated numerical values, a numerical value higher than the predetermined numerical value may be different from a numerical value not higher than the predetermined numerical value. Moreover, about the fashion item attribute relevant to the numerical value higher than the said predetermined numerical value, it is good also as a display mode different from the numerical value which is not higher than the said predetermined numerical value. By such different display, the viewer can easily understand the numerical value or the fashion item attribute which is a trend in the fashion before the cycle. In addition, rankings may be generated for the aggregated numerical values in descending order of numerical values. The numerical value or ranking may be displayed in descending order. The ranking may be a ranking for one or a plurality of specific fashion item attributes such as the type of fashion item, the color of the fashion item, and the pattern of the fashion item.

  Hereinafter, an image collected from another viewpoint and an example of application of the system of the present embodiment to this will be described. For example, fashion information related to a specific age group, fashion information related to a specific country or region, fashion information related to a person / group / object that he / she is interested in, fashion information related to what he / she frequently contacts explain.

<Fashion information related to specific age groups>
An image including a fashion item worn by a person of a specific age (hereinafter referred to as “specific age group image”) may be an image related to fashion worn by a person of a specific age, It may be an image relating to fashion worn by a person having a predetermined age range including age. The predetermined age range may be 2 years old, 3 years old, 5 years old, 10 years old, or the like. Different age ranges may be selected depending on the age. For example, a teenager may have a predetermined age range of 2 years old, a 20s may have a predetermined age range of 3 years, a 30s may have a predetermined age range of 5 years old, and the like. This is because, as the age increases, the relationship between age and appearance generally becomes milder.

  Part of the specific age group image may or may not overlap with part of the fashion image. In addition, images other than fashion images may or may not be included.

  The system flow of this embodiment when a specific age group image is used as the reference image is as shown in FIG. In FIG. 28, information on a specific age group is acquired in order to determine a specific age group image of the target image. Note that, as described above, the specific age group mentioned here may have a certain range or a specific age. The information on the specific age group may be selected in advance in the system according to the present embodiment, or the system according to the present embodiment may acquire information from a user terminal.

  In the system according to the present embodiment, when a specific age group image is used as a reference image, a “specific age group image attribute database” can be used instead of the fashion image attribute database when a fashion image is used.

  Even if the specific age group image attribute database is used, the system according to the present embodiment may display the same statistical process as described above. When the specific age group image attribute database is used in comparison with the case where the specific area is displayed in the fashion image attribute database, it is possible to display statistical processing based on information specific to each age.

  In addition, if the user is interested only in fashion for a specific age group, instead of buying or selling using the fashion image attribute database, the purchase destination for the fashion item specified with reference to the specific age group image attribute database This makes it possible to conduct business for those interested in a specific age group.

<Fashion information pertaining to a specific country or region>
An image including a fashion item related to a specific country or region (hereinafter referred to as “specific region image”) may be an image including a fashion item related to the specific country or region. The fashion item related to the specific country or region may be a fashion item having a cultural origin in the specific country or region, or a fashion item popular in the specific country or region. It may be a fashion item that is simply worn in the specific country or region. The region may be a plurality of countries or a part of a country.

  Part of the specific age image may or may not overlap with part of the fashion image. In addition, images other than fashion images may or may not be included.

  The system flow of this embodiment when a specific area image is used as a reference image is as shown in FIG. In FIG. 29, information on a specific area is acquired in order to determine a specific area image of the target image. As described above, the specific area mentioned here may be an area, a country, or a state or a municipality. The information on the specific area may be selected in advance in the system of the present embodiment, or the system of the present embodiment may acquire information from a user terminal.

  In the system according to the present embodiment, when a specific area image is used as a reference image, a “specific area image attribute database” can be used instead of the fashion image attribute database when a fashion image is used.

  Even if the specific area image attribute database is used, the system according to the present embodiment may perform the same statistical processing as described above and display the same. Compared with the case where the specific area is displayed in the fashion image attribute database, when the specific area image attribute database is used, it is possible to display statistical processing based on information unique to each area.

  In addition, if the user is interested only in fashion in a specific area, instead of buying or selling using the fashion image attribute database, refer to the purchase destination of the fashion item specified with reference to the specific area image attribute database. This enables business for those who are interested in a specific area.

<Fashion information for interested people / groups / objects>
In an image including a fashion item related to a user, group, or subject (hereinafter referred to as “interest image”), the person or group that the user is interested in may be a prominent person or group. It may be a person or group that is not prominent. It may be anime, movies, novels, paintings, etc. that you are interested in. For example, a fashion item related to a person who follows on Twitter, a fashion item related to a person who pressed “Like” or an article, a fashion item related to a celebrity who is a fan, and the like. An object of interest to the user may be, for example, a home page that is bookmarked or favorited on a PC used by the user.

  Part of the specific age image may or may not overlap with part of the fashion image. In addition, images other than fashion images may or may not be included.

  For example, FIG. 30 and FIG. 31 are shown as the system flow of the present embodiment when an interest image is used as a reference image, but is not limited thereto. In any case, in order to determine the specific interest image of the target image, information on the interest of the user may be acquired. FIG. 30 shows a method of searching for related information from the entire reference images in the system of the present embodiment based on information of interest based on information from the user terminal. This search method may be any method. When the information about the reference image is included, the reference image database may be in a format for selecting an image related to the user's interest from these. On the other hand, in FIG. 31, the system of the present embodiment holds an option of interest and an image corresponding to the option in advance, and presents the option of interest to the terminal used by the user. And the system of this embodiment may be a format which acquires the choice of interest from the terminal which a user uses, and uses the image corresponding to the selected interest as an interest image.

  In the system according to the present embodiment, when an interest image is used as a reference image, an “interest image attribute database” can be used instead of the fashion image attribute database when a fashion image is used.

  Even if the interest image attribute database is used, the system of the present embodiment may display the same statistical process as described above. When the interest image attribute database is used, it is possible to display statistical processing based on the user's specific interest information.

  In addition, when a user is interested in a specific person, group, or target, the user is interested by enabling the inquiry of the purchase place to the fashion item specified by referring to the interest image attribute database. Enable business with people, groups, or subjects.

<Fashion information related to contact>
In an image including a fashion item related to an object that the user often contacts (hereinafter referred to as “contact image”), the object that the user often contacts may be an object that the user is interested in. The target may not be the user's interest. Further, the degree of user contact may be determined using a predetermined standard. For example, in the case of a homepage that is browsed using a PC, a homepage that is browsed at a specific level or more among all the homepages that are browsed may be determined as a homepage that the user often contacts. The specific degree may be that the ratio of the browsing time to the total browsing time of the homepage is a predetermined value or more, or the ratio of the number of browsing times to the total number of browsing times of the homepage may be a predetermined value or more, These may be combined by a predetermined rule or weighting.

  A part of the contact image may or may not overlap with a part of the fashion image. In addition, images other than fashion images may or may not be included.

  The flow of the system of this embodiment when a contact image is used as a reference image is as shown in FIG.

  In the system according to the present embodiment, when a contact image is used as a reference image, a “contact image attribute database” can be used instead of the fashion image attribute database when a fashion image is used.

  Even when the contact image attribute database is used, the system according to the present embodiment may perform the same statistical processing as described above and display the same. When the contact image attribute database is used, it is possible to display statistical processing based on information that the user is specifically in contact with.

  In addition, it enables business related to the object that the user contacts by enabling the inquiry of the purchase place to the fashion item identified with reference to the contact image attribute database for the object that the user is in contact with. .

<Specific group-related fashion information>
An image (hereinafter referred to as “specific group image”) including a fashion item specified from the viewpoint of a specific designer or a plurality of designers or a specific one or a plurality of producers, sellers, and distributors of a fashion item. Fashion items designed by one or more fashion designers, fashion items designed by designers belonging to one or more design makers, prototype fashion items created by one or more producers / sellers, one or more productions It may be a fashion item scheduled to be sold by a seller / seller, a fashion item sold by one or more sellers, or a combination thereof.

  A part of the specific group image may or may not overlap with a part of the fashion image. In addition, images other than fashion images may or may not be included.

  The system flow of this embodiment when a specific group image is used as a reference image is as shown in FIG.

  In the system according to the present embodiment, when a specific group image is used as a reference image, a “specific group image attribute database” can be used instead of the fashion image attribute database when a fashion image is used.

  The above-mentioned fashion images, advanced fashion images, popular fashion images, specific age images, specific area images, interest images, contact images, specific group images and other images collected from a specific viewpoint (hereinafter referred to as “reference fashion images” ”, And the type of image collected from each viewpoint may be referred to as“ reference fashion image type. ”) May be collected by various methods.

  In the above description, the manner of collection relating to the reference image is equally applicable to these reference fashion images.

  Note that information that varies depending on the user, such as an interest image and a contact image, may be collected using the user's information. That is, the reference fashion image may be collected from the database of the system according to the present embodiment, the information processing apparatus on the network, and the like using the information of the PC related to the user. In addition, these pieces of information may be updated appropriately according to the time axis. Here, the PC related to the user is a PC owned by the user, a PC managed by the user, a PC for which the user has an account, a PC for which the user has an access right, or the like. Good. In addition, the PC information related to the user may use information of an application used by the user in the PC related to the user. For example, it may be information such as browsing history on the browser, bookmarks on the browser, and favorites.

  In addition, the system according to the present embodiment may have a configuration capable of selecting an information source as a reference image. For example, the system of the present embodiment may include a selection unit that selects a reference fashion image type.

  Note that the selection unit is not limited to selecting one of the plurality of reference fashion image types described above, and may be a unit that performs a set operation on a plurality of reference fashion image types. The set operation may be, for example, a sum, product, difference, complement set, and the like. Thus, for example, by calculating the sum of a plurality of specific age images and a plurality of specific area images, information processing regarding a fashion image of a specific age and a specific area becomes possible. Thus, by performing the statistical processing, the viewer can understand the tendency of the specific age and the specific area. In addition, by adopting a mode in which a set operation can be performed on the reference fashion image type, the user can acquire information on the desired reference fashion image type.

  Note that the target may be limited at the set stage, or the browsing target may be limited at the statistical processing stage. However, when the target is limited at the set stage, the target to which machine learning is applied decreases. Therefore, there is an advantage of reducing the amount of calculation and the storage amount of the database after applying machine learning. On the other hand, when the browsing target is limited at the stage of statistical processing, there is an advantage that information other than the target browsing target can be easily browsed only by the calculation amount of the statistical processing.

  In the present embodiment, description has been made using an attribute database by applying an attribute information generation apparatus using machine learning, but attribute information may be generated by other methods. For example, the attribute database may be generated by manual input, or may be appropriately modified by a rule-based method or manual input with respect to the original attribute database generated mechanically. An attribute database may be generated. It is conceivable that information collected in other databases is collected and different portions are generated by a rule base or the like. It is also conceivable that an updated part or the like of an image to be newly added to a database once generated by an attribute information generation function based on machine learning is manually input.

Therefore, the system of this embodiment does not have an attribute information generation function based on the machine learning function,
A database containing multiple attribute information;
A statistical processing unit for performing statistical processing;
May be provided.

  Moreover, the system of this embodiment does not have the attribute information generation function based on the machine learning function, and may have one or more functions that the above-described statistical processing unit can have.

  The system of this embodiment that does not have an attribute information generation function has an advantage of having the above-described various effects only by preparing an attribute database without burdens such as machine learning for the attribute information generation function.

  Furthermore, even if the system of this embodiment does not have the attribute database itself, it may be sufficient if the attribute database can be accessed and the attribute information can be obtained by other persons. This is because the attribute database may be available from companies that manage and provide the database.

Therefore, the system of this embodiment is
An acquisition unit for acquiring information including a plurality of attribute information via a network;
A statistical processing unit that performs statistical processing using the plurality of attribute information;
May be provided.

  Moreover, the system of this embodiment does not have an attribute database, and may have one or more functions that the above-described statistical processing unit can have.

  Further, the information including the plurality of attribute information may be information including each attribute value corresponding to the plurality of attribute information such as a record, or one of the attribute values such as a field. Information including attribute values may be a value collected in association with the plurality of attribute information.

3.3. Embodiment 3
Sometimes I want to know how the target fashion item is related to the fashion I refer to. In the following, the above-mentioned fashion image is used as the fashion for reference, but the advanced fashion image, popular fashion image, periodic fashion image, brand image, specific age image, specific area image, interest image, contact image, specific group It may be a reference fashion image such as an image or other images collected from a specific point of view, or may be a set of a plurality of reference fashion image types.

  The hardware configuration according to the system of the present embodiment may be the same as the hardware configuration described in the first embodiment, or may have different parts. For example, a description will be given with reference to FIG. The system according to this embodiment includes a system server 21. The system of the present embodiment may further include a machine learning device 22. Further, the system of the present embodiment may further include a product database 24. Further, the system of the present embodiment may include the terminal 23.

  In FIG. 34, although two machine learning devices 22a and 22b are shown, two or more machine learning devices or a single machine learning device may be used. The machine learning device may be a device that performs machine learning, or may have an attribute information generation function that has been generated by machine learning and has already been machine learning device. Even if the attribute information generation device has been machine-learned, machine learning may be performed using new learning data, and a function for performing such machine learning may be included.

  In FIG. 34, two product databases 24, 24a and 24b, are shown, but two or more product databases may be used as described above, or a single product database may be used.

  In FIG. 34, only one system server 21 is shown. However, as described above, the system server 21 may be composed of a plurality of information processing apparatuses, or a part of the system server 21 may be positioned via a network.

  In FIG. 34, the machine learning device 22 and the product database 24 are shown at positions via a network separately from the system server, but either one or both of the machine learning device 22 and the product database 24 are It may be included in the system server 21.

  The system server 21 may be connected to the machine learning devices 22a and 22b and the product databases 24a and 24b via the network 19 so that information can be transmitted to each other. Further, the system server 21 may be connected to the terminal 23 so that information can be mutually obtained.

  Further, the system of the present embodiment may have other hardware configurations. This is an example configuration without a machine learning device. Referring to FIG. 35, the system according to the present embodiment includes a system server 21. The system of the present embodiment may further include an attribute database 25. Further, the system of the present embodiment may further include a product database 24. Further, the system of the present embodiment may include the terminal 23.

  Although two attribute databases 25a and 25b are shown in FIG. 35, two or more attribute databases may be used as described above, or a single attribute database may be used. The attribute database has a database related to attribute information. The attribute information may be generated by an attribute information generation device or may be generated manually. When generated by the attribute information generation device, the attribute information generation device may be generated by machine learning, may be a rule-based method, or may be based on a function such as OCR from a medium. Good. The attribute database 25 may be connected to the system server 21 via the network 19, and the attribute database 25 may be included in the system server 21.

  The product database 24, system server 21, and terminal 23 in FIG. 35 may be the same as those in FIG.

  The system of this embodiment will be described with reference to FIG. First, a target image (hereinafter referred to as “analysis target image”) including a fashion item to be analyzed is prepared (step 1). The analysis target image may be, for example, an image obtained by capturing a fashion item held by a user with a digital camera, or an image displayed for sale on the net. .

  Next, the fashion item attribute which concerns on an analysis object image is produced | generated about the analysis object image using the attribute information generation function which the system of this embodiment has (step 2). The attribute information generation function may be generated by machine learning. For example, the attribute information generation function may be the attribute information generation function described in the first embodiment.

  Next, the analysis target image is compared with the data in the fashion image attribute database, and a fashion item that is the same as or similar to the fashion item in the analysis target image is specified in the fashion image attribute database (step 3). (Hereinafter, the specified fashion item is referred to as “specific item”). The image in the fashion image attribute database from which the specific item is extracted is referred to as “original fashion image”.

  The comparison between the analysis target image and the data in the fashion image attribute database may be a comparison between the fashion item attribute related to the fashion item in the analysis target image and the fashion item attribute in the fashion image attribute database. The comparison may compare attribute values in corresponding fashion item attributes.

  Identical and similar may be determined by various criteria. If the same and similar criteria are used as loose criteria, the number of subjects judged to be the same or similar will increase. If the same and similar criteria will be judged strictly, the criteria will be reduced. Is in a relationship. Below, although an example which determines the same or similar is demonstrated, it is not restricted to these, What is necessary is just a means which can determine the relationship of a fashion item as the same or similar. In particular, various methods may be used as long as they can be determined to be the same or similar using fashion item attributes.

  The same may be a case where the corresponding fashion item attributes are all the same, for example. For example, if the fashion item attribute includes information on the type of fashion item, the color of the fashion item, and information on the pattern of the fashion item, bottoms as the type of fashion item, black as the color of the fashion item, and floral pattern as the pattern of the fashion item The corresponding fashion item attribute attribute values such as.

  The similarity may be, for example, the case where the attribute value related to one fashion item attribute is the same among a plurality of corresponding fashion item attributes, but the attribute values related to other fashion item attributes are different. For example, the attribute value of the type of fashion item may be the same, but the attribute value of processing of the fashion item may be different.

  Moreover, in order to determine with similarity, the one or several fashion item attribute which should be the same may be defined. For example, the type of fashion item and the color of the fashion item may have the same attribute value relationship in order to be determined to be similar. The one or more fashion item attributes that should be the same may be determined in advance, or may be determined by being appropriately changed by machine learning described later.

  In this case, for the fashion item A and the fashion item B to be compared, when the fashion item A and the fashion item B are not the same, all the fashion items that the fashion items A and B have in common If it is determined that the corresponding fashion item attribute is the same attribute value for all of one or more fashion item attributes that should be the same among the attributes, the fashion item A and the fashion item B are similar. May be determined. Here, in addition, when one of the fashion items A and B has a fashion item attribute that the other does not have, it may not be determined as similar.

  In addition, a group is defined for the attribute value in the fashion item attribute in order to be determined to be similar, and the attribute value of the fashion item attribute relating to the two fashion items to be compared both belong to the related group May be determined to be similar. For example, in an outer of the type of fashion item, a coat and a trench coat are defined as related groups, and when one fashion item is a coat and the other fashion item is a trench coat, the one fashion item The other fashion items may be determined to be similar with respect to the type of the fashion item. On the other hand, in tops within the same fashion item type, for example, T-shirts and sweaters are defined as unrelated groups, where one fashion item is a T-shirt and the other fashion item is a sweater, One fashion item and the other fashion item may be determined not to be the same and similar with respect to the type of the fashion item.

  In this case, the fashion item A and the fashion item B to be compared are not the same as the fashion item A and the fashion item B, but all the fashion items A and B have in common If the item attributes are determined to be the same or similar, the fashion item A and the fashion item B may be determined to be similar. Here, in addition, when one of the fashion items A and B has a fashion item attribute that the other does not have, it may not be determined as similar.

  Similarities may be ranked in descending order of similarity of fashion item attributes instead of the same or similar fashion item attributes. For example, the similarity ranking may be set high for items having the same fashion item attribute, and the similarity ranking may be set according to the fashion item attribute, the above-described group in the fashion item attribute, or the like.

  The same or similar as described above can determine a combination of fashion item function item attributes that may be considered the same or similar when a person views the fashion item as the same or similar. It may be anything. These same or similar may be determined on a rule basis or may be determined by machine learning. For example, for the fashion item A and the fashion item B, the attribute relating to the fashion item A, the attribute relating to the fashion item B, and the result that the person has determined to be the same or similar to A and B are associated with each other. Alternatively, a machine learning device that has been subjected to machine learning may be used. The result determined that the person is the same or similar may be statistical data or a questionnaire result determined that the person is the same or similar. Moreover, although it described as the same or similar in the above-mentioned, only the same or similar may be sufficient.

<Display source>
The system of the present embodiment may acquire information relating to a fashion image for a specific item from a fashion image attribute database. Moreover, you may display the information which concerns on the acquired fashion image on a terminal.

  For example, the system of the present embodiment may display information relating to the origin of the fashion image as information related to the fashion image. The information on the source may be the name of the medium on which the image is published, the brand name, the magazine code, the ISBN code, the publisher, and / or the volume number.

  In the case where the analysis target image is an image of a fashion item held by the user, when the system of the present embodiment presents information on the source related to the specific item in the analysis target image, the viewer is It is possible to understand what kind of medium image the fashion item that is the same as or similar to the fashion item held by the user is displayed. At this time, if the medium is a famous or famous fashion magazine, the user understands that the fashion item held is the same or similar to the fashion item introduced in the famous or famous fashion magazine. In some cases, you can gain confidence in wearing your fashion items.

  In addition, when the analysis target image is, for example, an image obtained by imaging a fashion item possessed by another company, when the system of the present embodiment presents information on the source related to the specific item in the analysis target image, The viewer can understand what kind of medium images the same or similar fashion items held by other companies are displayed.

  In addition, when the analysis target image is an image on the Internet capturing a fashion item, when the system of the present embodiment presents information on the source related to the specific item in the analysis target image, the viewer It is possible to understand what kind of medium image the same or similar fashion item as the above fashion item is displayed. At this time, if the medium is a famous or famous fashion magazine, the user must confirm that the fashion item on the Internet is the same as or similar to the fashion item introduced in the famous or famous fashion magazine. Can deepen their interest in the fashion item.

<Display original fashion image>
In addition, when the system of the present embodiment presents an original fashion image as information relating to a fashion image, the viewer can see how the specific item similar to the fashion item possessed by the user is used. Since it can be visually seen, the specific item can be understood more. In particular, it can be understood with what other fashion items the specific item is used, and how it is used in a scene. To explain this from the user's perspective, a fashion item similar to his possession can have the opportunity to obtain information about other fashion items used in the fashion image.

<Display analysis target image and original fashion image>
In addition, when the system of the present embodiment displays the analysis target image and the original fashion image, the user can determine how similar the analysis target image and the original fashion image are, and how much they are different. Can be visually and intuitively understood.
In particular, when the system of the present embodiment displays the analysis target image and the original fashion image on the same screen, how much the analysis target image and the original fashion image are similar to each other and how much is different. It is possible for the user to understand intuitively on the same screen visually.

<Display of fashion item attributes of image to be analyzed and fashion item attributes of original fashion image>
Further, when the system of the present embodiment displays the information of the fashion item attribute related to the analysis target image and the information of the fashion item attribute related to the original fashion image, which fashion is the analysis target image and the original fashion image. Whether the item attributes are the same or which fashion item attributes are different can be understood not only as an intuitive display of an image but also as more accurate information. The information may be character string information. In the case of character string information, there is an advantage that the user can search on the Internet using the character string information.

  FIG. 37 is an example of the display described above. FIG. 37 shows an example in which the above-mentioned source display, analysis image object, and original fashion image are displayed. However, the display is not limited to this display example, and various displays may be performed. In addition, the fashion item attribute of the analysis target image and the fashion item attribute of the original fashion image may be displayed as information. Information may be displayed only when the fashion item attribute of the analysis target image is different from the fashion item attribute of the original fashion image. Also, different portions of the fashion item attribute and the fashion item attribute of the original fashion image may be displayed in a different form from those same portions.

  The system of the present embodiment further includes one or a plurality of fashion items in an original fashion image that is an image from which the specific item is extracted, and one or a plurality of fashion items other than the specific item ( Hereinafter, each fashion item of one or a plurality of fashion items other than the specific item may be referred to as “related specific item”) in the fashion image attribute database (step 4).

  The related specific item may be a fashion item other than the specific item among the fashion items associated with the original fashion image in the fashion image attribute database.

<Display related specific items>
The system of the present embodiment may display one or more related specific items. Moreover, the system of this implementation type | system | group may display the information of the fashion item attribute which concerns on a related specific item. FIG. 38 shows an example in which two related specific items are displayed. Moreover, it is an example which displays the information of a classification, a color, a pattern, and a process as a fashion item attribute which concerns on a related specific item. When these pieces of information are displayed, the viewer can search on the Internet independently using these pieces of information.

  When the analysis target image is an image of a fashion item held by the user and the system of the present embodiment displays the related specific item, the viewer combines with the fashion item held by the user. And can understand other fashion items. In fashion, there is a problem with the combination of fashion items, so the viewer can understand the fashion items that should be combined by displaying related specific items.

  In addition, when the analysis target image is an image obtained by capturing a fashion item held by the user and is an advanced fashion image as a reference image, when the system of the present embodiment displays a related specific item, The viewer can understand other fashion items that should be combined with the fashion items held by the user in the light of the advanced fashion image. Especially for viewers who like a combination of fashion in line with the latest fashion, there is an advantage of being able to refer to the latest fashion such as cutting-edge fashion images.

  In addition, when the analysis target image is an image of a fashion item held by the user and is a popular fashion image as a reference image, when the system of the present embodiment displays a related specific item, The viewer can understand other fashion items that should be combined with the fashion items held by the user in light of the popular fashion image. In particular, there is an advantage that a general fashion such as a popular fashion image can be referred to for a viewer who likes a combination of fashion that matches general fashion.

<Purchase from the product database>
Next, the system of the present embodiment uses a plurality of images (hereinafter referred to as “product images”) including fashion items that can be selected by the user using the attribute information generation function of the system of the present embodiment. For each product image (hereinafter referred to as “product database”), a fashion item attribute relating to the product image may be generated (step 5). The attribute information generation function may be generated by machine learning. For example, the attribute information generation function may be the attribute information generation function described in the first embodiment.

  Here, the product database may be provided in the system of the present embodiment, or may be stored in another information processing apparatus via the network. The other information processing apparatus may be a file server having an image or an information storage apparatus on the cloud. The owner of the product database may be different from the owner or administrator of the system of this embodiment.

  Further, the product database may store the product image and information on the fashion item related to the product image in association with each other. The information of the fashion item may be information such as a seller, manufacturer, price, and / or URL of the fashion item, for example. Information on these fashion items may be stored in association with information indicating a product image in a product attribute database described later.

  The selectable fashion item in the product image may be a fashion item that can be purchased or a borrowed fashion item.

  Next, the system of the present embodiment associates the information indicating the product image with the fashion item attribute related to the generated product image, and stores it in the storage unit as a product attribute database (step 6). ).

  Next, the system of the present embodiment compares one related specific item with a fashion item in a product attribute database, and the same or similar fashion item (hereinafter, “ "Related product") may be specified in the product attribute database (step 7). Here, the comparison, the same, and the similarity may be various methods as in the above-described technique. The comparison and similarity for specifying a specific item and the comparison and similarity for specifying a related-purpose product may be the same or different.

  The system of the present embodiment may present information related to one or a plurality of the related target products. Examples of the information related to the related-purpose product include the manufacturer of the related-purpose product, the sales start time of the related-purpose product, the price of the related-purpose product, and information related to the means for obtaining the related-purpose product. The information related to the related-purpose product obtaining means may be information related to the seller of the related-purpose product, information related to the rental agent of the related-purpose product, and the like. The information regarding the seller of the related-purpose product may be the location of the store of the related-purpose product, the URL related to the seller of the related-purpose product, and the like. The information related to one or a plurality of related purpose products may be stored and acquired in the product attribute database, or stored in another database associated with the related purpose products, and the information may be appropriately stored. The structure which can be acquired may be sufficient.

  FIG. 39 shows an example in which the seller name, the name of the product, the price of the product, and the related URL for online purchase are displayed as information related to the related-purpose product. The name of the product may be a model number. As for the price of goods, sale contents etc. may be indicated suitably. The related URL may be a URL that can be directly accessed for purchasing the product online, or a URL for describing the product. Although it is an example which displays two sellers about each related specific item, one seller may be sufficient and three or more sellers may be sufficient. Further, the number of sellers in the related specific item may be different.

<Identify or purchase fashion items that users are interested in>
The system of the present embodiment may be a system that can specify a selectable fashion item in the product attribute database that is the same as or similar to the specific item, instead of specifying the related specific item described above.

  This embodiment will be described with reference to FIG. First, a target image including a fashion item to be analyzed by the user is prepared as an analysis target image (step 1). The analysis target image may be, for example, an image including a fashion item owned by a user's friend or an image including a fashion item worn by a person longing for the user. These may be images found on the net or images taken with a digital camera or the like.

  Next, the fashion item attribute which concerns on an analysis object image is produced | generated about the analysis object image using the attribute information generation function which the system of this embodiment has (step 2). The attribute information generation function may be generated by machine learning. For example, the attribute information generation function may be the attribute information generation function described in the first embodiment.

  Next, the fashion item related to the analysis target image is compared with the fashion item in the product attribute database, and the same or similar fashion item as the fashion item related to the analysis target image is specified in the product attribute database ( Step 3). The identified fashion item is referred to as a “special purpose product”.

  Here, the comparison, the same, and the similarity may be various methods as in the above-described technique. The comparison and similarity for specifying a specific item and the comparison and similarity for specifying a related-purpose product may be the same or different.

  The system of the present embodiment may present information related to one or more specific purpose products. Examples of the information relating to the special purpose product include information on the manufacturer of the special purpose product, the sales start time of the special purpose product, the price of the special purpose product, and the means for obtaining the special purpose product. The information related to the means for obtaining the specific purpose product may be information about the seller of the specific purpose product, information about the rental company of the specific purpose product, and the like. The information related to the seller of the specific purpose product may be the location of the store of the specific purpose product, the URL related to the seller of the related purpose product, and the like. The information related to the one or more specific purpose products may be stored in the product attribute database and acquired, or stored in another database associated with the specific purpose products, and the information may be appropriately stored. The structure which can be acquired may be sufficient.

  The display example of the specific purpose product may be the same as the display example of the related purpose product shown in FIG.

  Also, in the system of the present embodiment, as described above, if there is attribute information, an attribute database, and / or a product attribute database related to the analysis target image, it is not necessary to have an attribute information generation function based on machine learning. .

The system of this embodiment is
An acquisition unit for acquiring one fashion item related to the analysis target image;
A specific unit that identifies the same or similar fashion item as the one fashion item in the attribute database using the fashion item attribute value;
May be provided.

The system of this embodiment is
An acquisition unit for acquiring a plurality of attribute values of one fashion item related to the analysis target image;
A plurality of attribute values, and a specification unit that identifies fashion items in which the attribute values corresponding to the plurality of attribute values have the same or similar attribute values in an attribute database;
May be provided.
3.4. Embodiment 4
The system of this embodiment creates a tendency of one image group and a tendency of another image group, and compares these tendencies.

  The system of this embodiment will be described with reference to FIG. First, one image group and another image group are prepared (step 1). Here, one image group is referred to as a “first company image database”, and the other image group is referred to as a “second company image database”. In the example of the present embodiment, the first company image database includes one fashion item manufactured and sold by the first company, and the second company image database is manufactured and sold by the second company. Includes one fashion item.

  Next, the fashion item attribute is generated for each image in the first company image database by using the attribute information generation function of the system of the present embodiment (step 2). The attribute information generation function may be generated by machine learning. For example, the attribute information generation function may be the attribute information generation function described in the first embodiment. The generated information may be stored. Here, the database generated based on the first company image database is referred to as a “first company image attribute database”. FIG. 42 is an example of the first company image attribute database. In association with each image ID, specific classifications about the type of fashion item, the color of the fashion item, and the like are stored as the fashion item attributes. Although the image ID is described here, the image ID may or may not be used when used only for the purpose of counting each attribute. In addition, an ID for simply distinguishing each information may or may not exist.

  Similarly, a fashion item attribute is generated for each image in the second company image database using the attribute information generation function of the system of the present embodiment (step 3). Here, the database generated based on the second company image database is referred to as “second company image attribute database”.

  Next, the specific attribute information is statistically processed for the first company image attribute database and the second company image attribute database (step 4). For example, FIG. 43 is an example of tabulating with the same attribute value of the type of fashion item. More specifically, as the type of fashion item, the numerical value of the total result is stored for each attribute value such as tops, outer, pants, skirt, and the like. The aggregation may be calculated so that the total number of fashion items or reference images having the same attribute value can be calculated. In FIG. 43, the aggregation result based on the first company image attribute database and the aggregation result based on the second company image attribute database are stored in the same database, but these are stored in separate databases. It is also possible to omit the step of storing in the database after counting and proceed with the display process described below.

  Next, the statistical processing unit may display the aggregation results in comparison. For example, FIG. 44 is an example in which the types of fashion items are tabulated and displayed for the first company product and the second company product. The types of fashion items were tops, outerwear, pants, skirts, bags and shoes. By the display of the system of the present embodiment, for example, the product of the fashion item of the first company and the product of the fashion item of the second company are compared in terms of the attribute values of the types of fashion items such as tops and outerwear. Can be displayed. This is because, when the classification method of the product type is not necessarily the same in each company, the attribute value related to the fashion item attribute generated through the attribute information generation function of the system of the present embodiment is used, so that the unified judgment standard Thus, it is possible to display. For example, such a display is a material for the first company to examine the difference in the product type of the product group of the second company, and can be used for marketing research and research of other companies.

  Further, the fashion attributes may be aggregated and displayed for each finer type of fashion item type. For example, FIG. 45 shows an example in which tops are targeted and T-shirts, cut-and-sew, shirts, blouses, polo shirts, sweaters, and the like are targeted. In this way, when the classification is particularly fine, the naming method of the product classification may be different for each company. For example, the judgment criteria for blouse and shirt may be unclear. When the attribute information generation device is used, it can be classified by uniform judgment according to the judgment criteria by the device, so that the number of product types of each company can be judged according to a certain standard, for example, the first company or the third Even a person or the like can conduct marketing and product surveys.

The system of this embodiment is
An image for machine learning, a plurality of attribute information of a plurality of fashion items included in the image for machine learning and / or attribute information of a person included in the image for machine learning are used, and , A neural network unit that machine-learns the relationship between fashion item attribute information and / or person attribute information by deep learning,
An acquisition unit for acquiring a first target image belonging to the first group;
By applying the first target image to the neural network, the first target image is composed of a plurality of attribute information of a plurality of fashion items related to the first target image and / or attribute information of a person related to the first target image. A generator for generating a fifth database;
By applying a second target image belonging to the second group to the neural network, a plurality of attribute information of a plurality of fashion items related to the second target image and / or attribute information of a person related to the second target image A generator for generating a sixth database comprising:
The system may include a statistical processing unit that statistically processes the information in the fifth database and the information in the sixth database using the fifth database and the sixth database. .

  The images belonging to the first group and the images belonging to the second group may be the various images described above. For example, the images belonging to the first group may be a plurality of images related to the first group, and the images belonging to the second group may be a plurality of images related to the second group. The first organization and the second organization may be different organizations.

  An example of the plurality of images related to the first group may be the above-mentioned first company image database, and an example of the plurality of images related to the second group may be the above-mentioned second company image database, An example of the database may be the aforementioned first company image attribute database, and an example of the sixth database may be the aforementioned second company image attribute database. These may be reversed.

  The statistical processing unit may be configured to generate and display a total numerical value for the first attribute information in the fifth database and a total numerical value for the first attribute information in the sixth database.

  By generating and displaying the total numerical value for the first attribute information, it becomes easy to compare the images belonging to the first group and the images belonging to the second group with respect to the first attribute information. In particular, by using the attribute information generation function that has already been machine-learned, it is possible to display the aggregate information about the first attribute information in the standard of the attribute information generation function, so that the viewer can easily compare with the same standard It becomes.

  In addition, a total value for the first attribute information in the fifth database and a total value for the first attribute information in the sixth database may be generated. Since the information generated by the configuration can be displayed by transmitting it through a network in any information processing apparatus, there is an advantage that it can be displayed even for a distant viewer.

  In the above description, the configurations of the fifth database and the sixth database have been described for the images belonging to the first group and the images belonging to the second group, respectively. It may be possible to construct a database. In this case, by applying the images belonging to each group to the neural network, a database composed of a plurality of attribute information of a plurality of fashion items and / or attribute information of a person is generated, and the first attribute information May be configured to aggregate and / or display. With this configuration, it is possible to compare images related to three or more groups.

  As described above, the attribute database may be generated by other means without using the attribute information generation function.

  Therefore, the system of the present embodiment includes a first database including a plurality of fashion item attributes and / or a plurality of person attributes and an eighth database including a plurality of fashion item attributes and / or a plurality of person attributes, respectively. It may be configured to generate a total numerical value for the attribute information. Further, the generated total numerical value may be displayed in comparison.

  Further, the system according to the present embodiment may be configured such that, in the configuration described above, the seventh database and the eighth database described above are stored in another information processing device or storage device, and a part thereof is acquired from them.

  In addition, the above two databases, the seventh database and the eighth database, are not limited to using three or more databases or acquiring information from these databases, It may be configured to be generated and / or displayed.

  Further, in the above description, as a single image group or another image group, the description has been mainly focused on specific group-related images. However, as one image group or another image group, a plurality of the above-described reference fashion images, That is, even if a plurality of images such as advanced fashion images, popular fashion images, periodic fashion images, brand images, specific age images, specific area images, interest images, contact images, and other images collected from a specific viewpoint are prepared. Alternatively, a set operation may be performed on a plurality of reference fashion image types.

The result of the statistical processing may be displayed according to the selected one image group and the other selected image group.
3.5. Embodiment 5
The system of the present embodiment specifies a combination of fashion items in one image group with reference to information in another image group. For example, the system according to the present embodiment identifies an appropriate fashion item when combined with the one fashion item with respect to one fashion item in the fashion item that the user has, based on the reference image, and is identified from the fashion item that the user has. It is a system that can.

  The system of this embodiment will be described with reference to FIG. The system according to the present embodiment includes a system server 21 and an attribute database 25, and may be configured to acquire information input from a terminal.

  Here, the attribute database 25a is, for example, an advanced fashion image attribute database related to advanced fashion images, and the attribute database 25b is, for example, an attribute database related to fashion items possessed by a user, referred to as a user attribute database. I will do it.

  As described above, the attribute database may be configured from attribute information generated using the machine learning function, may be configured from attribute information generated by other methods, or may be information input by a human being. May be composed of attribute information based on Moreover, the information regarding the fashion item concerning attribute information may be stored in association with the attribute information.

  Next, the system flow of the present embodiment will be described with reference to FIG. First, a plurality of attribute information related to one fashion item is acquired from the user attribute database and stored in the storage unit (step 1).

  Next, the same or similar fashion item as the one fashion item is specified from the advanced fashion image attribute database using the attribute information (step 2). In the example of the system of the present embodiment, the identified fashion item is referred to as a “specific item”.

  Next, in the advanced fashion image attribute database, another fashion item of the advanced fashion image related to the specific item is specified (step 3). In the example of the system of this embodiment, the identified fashion item is referred to as a “related item”. Here, the related item may be one or plural.

  Next, one or a plurality of related items and the same or similar fashion items are specified in the user attribute database (step 4).

  By adopting the above-described configuration, the system according to the present embodiment can identify a fashion item to be combined in the light of an advanced fashion image from among a plurality of fashion items possessed by the user. There is an advantage that the combination of

  One image group or another image group includes a plurality of the above-mentioned reference fashion images, that is, advanced fashion images, popular fashion images, brand images, specific age images, specific area images, interest images, contact images, specific group related A plurality of images such as images and other images collected from a specific point of view may be prepared, or a set operation may be performed on a plurality of reference fashion image types.

  As an example of the system of the present embodiment, for example, an attribute database related to one image group is an attribute database related to stage costumes, and an attribute database related to fashion images of a specific era / region is used as an attribute database related to other image groups; Then, it is possible to identify other fashion items that should be combined with one fashion item in the stage costume in the light of the fashion images of a specific era / region. You can select a combination of fashion items according to the times and regions. For example, if the stage costume is already owned by the user, it is possible to specify a better combination of fashion items among the held fashion items. In addition, if the attribute database relating to the stage costume is, for example, a product attribute database that can be purchased or ordered, fashion items in line with the specific period / region described above can be purchased or ordered. Can be selected, and a fashion item having an atmosphere more in line with the specific period / region can be specified. Moreover, in the system of this embodiment, the above-mentioned specific or selected fashion item may be generated as information, and may be displayed so that it can be viewed by the user.

  In the above description, the image may be image data or an image file. Various data formats or data formats may be used. The image may be in a raster format or a vector format. Examples include, but are not limited to, BMP, GIF, JPEG, PNG, PNM, TIFF, XPM, WebP, PS, EPS, PDF, SVG, WMF, and the like.

  The image may be a part of a moving image. For example, the image may be a still image that is a part of a moving image.

  In the above description, the configuration implemented by the system of this example has been described. However, these configurations may be implemented by one or more information processing apparatuses in the system.

  The invention examples described in the embodiments of the present application are not limited to those described in the present application, and may be applied to various examples within the scope of the technical idea. For example, in the system according to the embodiment of the present document, information presented on the screen of one information processing apparatus can be transmitted to the other information processing apparatus for display on the screen of the other information processing apparatus. The system of each embodiment may be configured such that the one information processing apparatus has a transmission unit.

  Further, the processes and procedures described in this application document may be realized not only by those explicitly described in the embodiment but also by software, hardware, or a combination thereof. In addition, the processes and procedures described in the present application may be implemented as a computer program and executed by various computers.

DESCRIPTION OF SYMBOLS 10 ... Information processing device 11 ... Bus 12 ... Arithmetic device 13 ... Storage device 14 ... Input device 15 ... Display device 16 ... Communication IF
DESCRIPTION OF SYMBOLS 19 ... Network 21 ... System server 22 ... Machine learning device 23 ... Terminal 24 ... Product database 25 ... Attribute database 26 ... Product attribute database

Claims (29)

For fashion images including fashion images related to domestic shop clerk,
An image for machine learning and one or more attribute information included in the image for machine learning are used, and the attribute information includes a plurality of fashion item attributes in each of the image and the plurality of fashion items, An acquisition unit that acquires attribute information generated by applying a machine learning unit that machine-learned the relationship of
An acquisition unit for acquiring information relating to the fashion image;
An associating unit that generates an element of the first database configured by associating the acquired attribute information and information related to the fashion image;
A statistical processing unit for performing statistical processing;
A system comprising a,
The information related to the fashion image includes age information of a predetermined person in the fashion image.
System . The statistical processing unit displays one or a plurality of attribute information to be displayed in time series for the user to select, and generates a total number for each selected attribute information.
The system of claim 1. The statistical processing unit generates information on a ratio of one color relative to the whole color in a predetermined period.
The system according to claim 1 or 2. The statistical processing unit generates information on a proportion of one or more fashion item attributes relative to the whole,
The system according to any one of claims 1 to 3. The statistical processing unit generates numerical information indicating a relationship between a numerical value of one or more fashion item attributes in one year and a numerical value of the one or more fashion item attributes in a year before the one year. ,
The system according to any one of claims 1 to 4. The statistical processing unit displays one or a plurality of attribute information for selection by a user, and only for data having the selected one or a plurality of attribute information as attribute information, for each specific attribute information or Generate a total number for each selected attribute information,
The system according to any one of claims 1 to 5. The information related to the fashion image includes a brand related to the fashion image.
The system according to any one of claims 1 to 6. The information relating to the fashion image includes the gender of a pre-given person in the fashion image,
The system according to any one of claims 1 to 7. The fashion image is included in one selected image database of a plurality of image databases each including a plurality of image data.
The system according to any one of claims 1 to 8. The system
For the acquired second target image, an image for machine learning and one or a plurality of attribute information included in the image for machine learning are used, and the relationship between the image and the attribute information is determined by a machine. A specifying unit for specifying, in the first database, the second fashion item associated with the attribute information generated by applying the learned machine learning unit and the same or similar attribute information;
The system according to any one of claims 1 to 9. The second target image is an image obtained by capturing a fashion item held by a user.
The system according to claim 10. The system
For the acquired third target image including a selectable fashion item, an image for machine learning and one or a plurality of attribute information included in the image for machine learning are used, and an image, Attribute information generated by applying a machine learning unit that machine-learned the relationship between the attribute information,
Information related to the third target image;
An associating unit for associating and generating elements of the second database;
The system
A third fashion item that is the same as or similar to the second fashion item is provided in the second database.
The system according to claim 10. The system
For the acquired fourth target image including a selectable fashion item, an image for machine learning and one or a plurality of attribute information included in the image for machine learning are used, and an image, Attribute information generated by applying a machine learning unit that machine-learned the relationship between the attribute information,
Information on the fourth target image;
An associating unit for associating and generating elements of the second database;
The system
A specifying unit for specifying, in the first database, a fourth fashion item different from the second fashion item in the image relating to the second fashion item;
A specifying unit that specifies, in the second database, a fifth fashion item that is the same as or similar to the fourth fashion item;
The system of claim 10. A machine learning image and a plurality of attribute information included in the machine learning image are used for each of a plurality of images including a fashion item related to a user, and the image, the attribute information, , Attribute information generated by applying a machine learning unit that machine-learned the relationship between
A fourth database associated with each of the images;
A specifying unit that specifies, in the first database, a sixth fashion item that is the same as or similar to the one fashion item related to the user;
A specifying unit for specifying a seventh fashion item different from the sixth fashion item in the image relating to the sixth fashion item;
A specifying unit for specifying an eighth fashion item identical or similar to the seventh fashion item in the fourth database;
Comprising
The system according to any one of claims 1 to 13. For the acquired second image,
Machine learning image and one or more attribute information included in the machine learning image are used and generated by applying a machine learning unit that machine-learns the relationship between the image and the attribute information. Using the attribute information, a specifying unit that specifies, in the first database, a second fashion item that is the same as or similar to the fashion item according to the second image;
For reference images that contain acquired fashion items,
Machine learning image and one or more attribute information included in the machine learning image are used and generated by applying a machine learning unit that machine-learns the relationship between the image and the attribute information. Attribute information,
A second database associated with information relating to a reference image including the selectable fashion item;
A specifying unit that specifies, in the second database, a third fashion item that is the same as or similar to the second fashion item;
Comprising
15. The system according to any one of claims 1 to 14. For each fashion image including a fashion image related to a domestic shop clerk and a fashion item related to the acquired first group,
An image for machine learning and one or more attribute information included in the image for machine learning are used, and the attribute information includes a plurality of fashion item attributes in each of the image and the plurality of fashion items, An acquisition unit that acquires attribute information generated by applying a machine learning unit that machine-learned the relationship of
An acquisition unit for acquiring information relating to the fashion image;
An association unit that associates the acquired attribute information with the information related to the fashion image to generate an element of the first database;
For each image containing fashion items related to the acquired second group,
Machine learning image and one or more attribute information included in the machine learning image are used and generated by applying a machine learning unit that machine-learns the relationship between the image and the attribute information. Attribute information,
Information relating to the fashion image;
A sixth database associated with
A statistical processing unit for statistically processing the information in the first database and the information in the sixth database;
A system comprising :
The information related to the fashion image includes age information of a predetermined person in the fashion image.
System . The information relating to the fashion image includes information relating to the source of the fashion image,
The system according to any one of claims 1 to 16. The information relating to the fashion image includes the time when the fashion image was published,
The system according to any one of claims 1 to 17. Including memory,
The system according to any one of claims 1 to 18. Including arithmetic units,
The system according to any one of claims 1 to 19. Computer
For a fashion image including a fashion image related to a domestic shop clerk, an image for machine learning and one or more attribute information included in the image for machine learning are used, and the image and a plurality of fashion items Obtaining attribute information generated by applying a machine learning unit that machine-learned the relationship between attribute information including a plurality of fashion item attributes in
An acquisition step of acquiring information relating to the fashion image;
An association step of associating the acquired attribute information with information about the fashion image to generate an element of the first database;
For the acquired second target image, an image for machine learning and one or a plurality of attribute information included in the image for machine learning are used, and a relationship between the image and the attribute information is determined by a machine. Identifying, in the first database, the second fashion item associated with the attribute information generated by applying the learned machine learning unit and the same or similar attribute information;
Is a method of performing
The information related to the fashion image includes age information of a predetermined person in the fashion image . In the method, the computer
For the acquired third target image including a selectable fashion item, an image for machine learning and one or a plurality of attribute information included in the image for machine learning are used, and an image, Attribute information generated by applying a machine learning unit that machine-learned the relationship between the attribute information,
Information related to the third target image;
Associating and generating a second database element;
Identifying in the second database a third fashion item identical or similar to the second fashion item;
The method of claim 21, wherein: In the method, the computer
For the acquired fourth target image including a selectable fashion item, an image for machine learning and one or a plurality of attribute information included in the image for machine learning are used, and an image, Attribute information generated by applying a machine learning unit that machine-learned the relationship between the attribute information,
Information on the fourth target image;
Associating and generating a second database element;
Identifying in the first database a fourth fashion item different from the second fashion item in the image relating to the second fashion item;
Identifying a fifth fashion item in the second database that is the same as or similar to the fourth fashion item;
The method of claim 21 or 22, wherein: Computer
An image for machine learning and one or a plurality of attribute information included in the image for machine learning are used for a fashion image including a fashion image related to the acquired domestic shop clerk. Obtaining attribute information generated by applying a machine learning unit that machine-learned the relationship between attribute information including each of a plurality of fashion item attributes in the fashion item;
An acquisition step of acquiring information relating to the fashion image;
An association step of associating the acquired attribute information with the information relating to the fashion image to generate an element of the first database;
A machine learning image and a plurality of attribute information included in the machine learning image are used for each of a plurality of images including a fashion item related to a user, and the image, the attribute information, , Attribute information generated by applying a machine learning unit that machine-learned the relationship between
An association step, wherein the images are associated with each other to generate a fourth database element;
Identifying a sixth fashion item in the first database that is the same as or similar to the one fashion item associated with the user;
Identifying a seventh fashion item different from the sixth fashion item in the image relating to the sixth fashion item;
Identifying an eighth fashion item in the fourth database that is the same as or similar to the seventh fashion item;
Is a method of performing
The information related to the fashion image includes age information of a predetermined person in the fashion image . Computer
Against the fashion image according to the acquired domestic shop clerk,
An image for machine learning and one or more attribute information included in the image for machine learning are used, and the attribute information includes a plurality of fashion item attributes in each of the image and the plurality of fashion items, An acquisition step of acquiring attribute information generated by applying a machine learning unit that machine-learned the relationship of
Obtaining information relating to the fashion image;
An association step of associating the acquired attribute information with information relating to the fashion image to generate an element of the first database;
For the acquired second image,
Machine learning image and one or more attribute information included in the machine learning image are used and generated by applying a machine learning unit that machine-learns the relationship between the image and the attribute information. Identifying in the first database a second fashion item that is the same as or similar to the fashion item according to the second image, using the attribute information that has been created;
For reference images that contain acquired fashion items,
Machine learning image and one or more attribute information included in the machine learning image are used and generated by applying a machine learning unit that machine-learns the relationship between the image and the attribute information. Attribute information,
Associating information relating to a reference image including the selectable fashion item to generate an element of the second database;
Identifying in the second database a third fashion item identical or similar to the second fashion item;
Is a method of performing
The information related to the fashion image includes age information of a predetermined person in the fashion image . The computer includes a memory;
26. A method according to any one of claims 21 to 25. The computer includes a computing device;
27. A method according to any one of claims 21 to 26.   The program for functioning a computer as a system as described in any one of Claims 1 thru | or 20.   A program for causing a computer to execute the method according to any one of claims 21 to 27.

Images