KR20230030870A - Apparatus and method for apparel recommendation, computer-readable storage medium and computer program - Google Patents

Abstract

According to the embodiment of the present invention, a clothing recommendation device includes: a clothing vector collector which collects clothing vector information; a stylist module which configures the clothing vector information and the pre-collected style vector information as a coordinate vector, is trained by input of the coordinate vector; and a clothing recommendation unit which enters any one of the user's clothing information and one or more of the style information selected by the user in the stylist module, and recommends the clothing to the user based on the clothing information output from the stylist module. An embodiment of the present invention can effectively recommend clothing which considers user tastes by using the stylist module, which is a deep learning model.

Description

Apparatus and method for recommending clothing, computer readable recording medium and computer program

Embodiments relate to a clothing recommendation apparatus and method for recommending clothing.

In general, clothing has functions as a means of expressing the person in addition to essential functions such as maintaining body temperature and protecting skin. Protect your body by wearing clothes to adapt to climate change, and express your individuality by wearing clothes that suit you. When the function of these clothes is effectively displayed in everyday clothes, it is possible to lead a physically pleasant and psychologically satisfactory clothing life.

There are countless combinations of clothes, and finding a combination that suits me in the flood of clothes is a pleasure for many people, but at the same time, it is often a problem.

As a conventional clothing recommendation method, coordinating services can be received remotely by utilizing a communication network, but intervention by a third party or a professional coordinator is essential for coordinating services.

Accordingly, it is difficult to provide the style desired by the user in that it takes considerable time to reach the coordinating service, additional costs such as labor costs are incurred, or the criteria for judging the coordinating service are different.

In order to solve the above problems, an object of the embodiment is to provide a clothing recommendation apparatus and method for recommending clothes and coordination in consideration of a style desired by a user.

An apparatus for recommending clothing according to an embodiment includes: a clothing vector collection unit that collects clothing vector information; a stylist module configured to configure the clothing vector information and pre-collected style vector information as coordinating vectors and trained using the coordinating vectors as input; and a clothing recommendation unit that inputs at least one of user clothing information and style information selected by the user to the stylist module and recommends clothing to the user based on the clothing information output from the stylist module.

The clothing vector collection unit may include a clothing information collection unit that extracts clothing information from pre-collected clothing images using an artificial neural network, and a clothing information encoding unit that converts the extracted clothing information into the clothing vector information in a vector form. can

The clothing vector collection unit includes a clothing information collection unit that extracts clothing information by matching pre-collected product tags and product information with a clothing information database, and converts the extracted clothing information into the clothing vector information in a vector form. An encoding unit may be included.

The stylist module configures coordinating vector information with no information loss among the clothing vector information and pre-collected style vector information, performs primary training, and performs primary training, and performs some information among the clothing vector information and pre-collected style vector information. Secondary training can be performed by composing the lost vector information into a coordinating vector.

The clothing recommendation unit may enter a coordination recommendation mode or a product recommendation mode based on input information selected by the user.

In the coordination recommendation mode, the clothing recommendation unit may recommend clothing by matching clothing information output from the stylist module with clothing information possessed by the user.

In the product recommendation mode, the clothing recommendation unit may recommend clothing by matching clothing information output from the stylist module with clothing information retrieved from an external clothing product DB.

The clothing recommendation method of the embodiment includes the steps of collecting clothing vector information, constructing the clothing vector information and pre-collected style vector information into coordinating vectors, and training a stylist module by using the coordinating vectors as inputs; and inputting at least one of clothing information and style information selected by the user to the stylist module, and recommending clothing to the user based on the clothing information output from the stylist module.

The collecting of clothing vector information may include extracting clothing information from pre-collected clothing images using an artificial neural network, and converting the extracted clothing information into the clothing vector information in a vector form. there is.

The collecting of clothing vector information may include extracting clothing information by matching product tags and product information collected in advance with a clothing information database, and converting the extracted clothing information into the clothing vector information in a vector form. can include

In the step of training the stylist module, vector information in which information is not lost among the clothing vector information and pre-collected style vector information is configured as a coordinating vector for primary training, and the clothing vector information and pre-collected style vector information are firstly trained. Secondary training can be performed by constructing vector information in which some of the information is lost as a coordinating vector.

In the clothing recommendation step, a coordination recommendation mode or a product recommendation mode may be entered based on input information selected by the user.

In the coordinating recommendation mode, in the step of recommending clothing, clothing information output from the stylist module may be matched with clothing information owned by the user to recommend clothing.

In the product recommendation mode, in the step of recommending clothes, clothes may be recommended by matching clothes information output from the stylist module with clothes information retrieved from an external clothes product DB.

In addition, the embodiment is a computer readable recording medium storing a computer program, wherein the computer program, when executed by a processor, includes the steps of collecting clothing vector information, and the clothing vector information and pre-collected style vector information. constructing a coordinating vector and training a stylist module using the coordinating vector as an input, inputting at least one of the user's clothing information and the user's selected style information to the stylist module, and from the stylist module It may include instructions for causing the processor to perform an operation including recommending clothing to the user based on the output clothing information.

In addition, the embodiment is a computer program stored on a computer-readable recording medium, and the computer program, when executed by a processor, includes the steps of collecting clothing vector information, and the clothing vector information and pre-collected style vector information. constructing a coordinating vector and training a stylist module using the coordinating vector as an input, inputting at least one of the user's clothing information and the user's selected style information to the stylist module, and from the stylist module It may include instructions for causing the processor to perform an operation including recommending clothing to the user based on the output clothing information.

The embodiment has an effect of effectively recommending clothing considering user taste by using a stylist module, which is a deep learning model.

In addition, the embodiment has an effect of enabling objective and consistent clothing recommendation by using a deep learning model.

In addition, the embodiment has an effect of obtaining various clothing recommendation information by using big data.

1 is a block diagram illustrating a clothing recommendation system according to an embodiment.
2 is a block diagram illustrating an apparatus for recommending clothes according to an embodiment.
FIG. 3 is a diagram illustrating a clothing vector collection unit of FIG. 2 .
4 is a diagram showing the structure of a coordinating vector according to an embodiment.
5 is a view showing how the stylist module of FIG. 2 is trained.
FIG. 6 is a diagram illustrating a clothing recommendation unit of FIG. 2 .

Hereinafter, embodiments will be described in detail with reference to the drawings.

1 is a block diagram showing a clothing recommendation system according to an embodiment, FIG. 2 is a block diagram showing a clothing recommendation device according to an embodiment, FIG. 3 is a diagram showing a clothing vector collection unit of FIG. 2, and FIG. A diagram showing the structure of a coordinating vector according to an example, FIG. 5 is a diagram showing how the stylist module of FIG. 2 is trained, and FIG. 6 is a diagram showing the clothing recommendation unit of FIG. 2 .

Referring to FIG. 1 , a clothing recommendation system according to an embodiment may include a clothing recommendation device 100 , a server 200 , and a user terminal 300 .

The clothing recommendation device 100 may recommend clothing by analyzing information on clothing images collected from the server 200 and the user terminal 300 . Recommended information may be provided to the user through the user terminal 300 .

The server 200 may be a server that provides data such as clothing information and style information to the clothing recommendation device 100 . The server 200 may include a web server or an application server, but is not limited thereto.

The user terminal 300 may provide information such as clothing owned by the user to the clothing recommendation device 100 . The user terminal 300 includes a desktop computer operated by a user, a smart phone, a laptop computer, a tablet PC, a smart TV, a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, It may be an e-book reader, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, and other mobile or non-mobile computing devices, but is not limited thereto.

In addition, the user terminal 300 may be a wearable terminal having a communication function and data processing function, such as a watch, glasses, a hair band, and a ring. The user terminal 300 is not limited to the above, and a terminal capable of web browsing may be borrowed without limitation.

As shown in FIG. 2 , the clothing recommendation device 100 may include a clothing vector collection unit 110 , a stylist module 120 , and a clothing recommendation unit 130 .

The clothing vector collecting unit 110 may convert clothing information into clothing vector information.

As shown in FIG. 3 , the clothing vector collecting unit 110 may include a clothing information collecting unit 112 and a clothing information encoding unit 114 .

The clothing information collecting unit 112 may extract clothing information from the previously collected clothing image 10 using the artificial neural network 111 . Clothing information may include a table form in which items and values exist. An item may include a type, color, pattern, and the like, and a value may include a value for the item.

Unlike this, the clothing information collecting unit 112 may collect clothing information by matching pre-collected product tags with the clothing information DB 30 using a matching module. Here, the clothing information DB 30 may be included in the server 200 .

The clothing information encoding unit 114 may convert clothing information extracted from the clothing information collecting unit 112 into clothing vector information 115 in a vector form. The clothing information encoding unit 114 may encode clothing information into a low-dimensional vector using an artificial neural network having an autoencoder structure or an embedding technology such as word2vec.

The clothing vector information 115 may include tops (inner and outer) and bottoms (inner and outer) from clothing information.

Returning to FIG. 2, the stylist module 120 configures clothing vector information and pre-collected style vector information as coordinating vectors, and deep learning with a denoising autoencoder structure trained using the coordinating vectors as inputs can be a module.

As shown in FIG. 4 , the coordinating vector 122 may include clothing vector information 115 and style vector information 121 . Here, the style vector information 121 is a vector representing the style of the overall costume coordination, and may be information expressed as a vector by digitizing a theme (casual, street, vintage, etc.), color tone, and season. The style vector information 121 can be obtained through one-hot encoding, word embedding, or the like according to the classification of themes, color tones, and seasons. For example, the style vector information 121 based on one-hot encoding may be a vector filled with 1 for a corresponding theme and 0 for others. The style vector information 121 is information in which style information prepared by an external expert is divided into a vector form, and may be previously collected from the external server 200 .

Steps for training the stylist module 120 are as follows.

As shown in FIG. 5 , the stylist module 120 may receive a coordinating vector 122 from a coordinating vector DB 123 . The coordinating vector DB 123 is a DB that stores the coordinating vectors 122 and may be included in the server 200 .

The stylist module 120 may be trained using the coordinating vector 122 as an input. Here, the input coordinating vector may be a coordinating vector in which information is not lost. The stylist module 120 may be trained in a direction of reducing a restoration error with an output coordinator vector, and for this purpose, the stylist module 120 may be configured as a backpropagation neural network model.

The stylist module 120 may be trained by using an input coordinator vector generated as the coordinator vector 122 in which some information is lost as an input. Areas in which some information is lost may be filled with values such as 0 or -1. The stylist module 120 may restore lost information using an input coordinating vector in which some information is lost.

The stylist module 120 can be trained in two stages using an input coordinator vector with no information loss and an input coordinator vector with information loss. Alternatively, the stylist module 120 may perform training using only the input coordinator vectors in which information is lost.

Returning to FIG. 2 , the clothing recommendation unit 130 may recommend clothing in consideration of user preferences.

As shown in FIG. 6 , the clothing recommendation unit 130 outputs an output coordinating vector using the stylist module 120 by using a part of the clothing information and style information input by the user as input, and outputs the output coordinating vector Based on this, clothing can be recommended to the user.

When specific clothing is determined, the user may input corresponding clothing information (eg, jeans), and the clothing recommendation unit 130 may convert the clothing information into a clothing encoding vector and input the corresponding part of the coordinating vector. If specific clothing is not determined, the clothing encoding vector may be filled with a value representing no information.

When a user inputs a desired style (theme, color, season, etc.), the clothing recommendation unit 130 may generate a style vector according to the selected style. When the user does not select a style, the style vector may be filled with a random value or a value representing no information.

That is, the input coordinating vector input to the stylist module 120 may be generated by combining the clothing encoding vector according to the user's input and the style vector.

As described above, the clothing recommendation unit 130 may recommend clothing based on the output coordinating vector output through the stylist module 130 when the stylist module receives the coordinating vector into which user information is input.

If the user desires either a coordinating recommendation mode or a product recommendation mode, the clothing recommendation unit 130 may proceed to the coordinating recommendation mode or product recommendation mode based on user input information.

In the coordinating recommendation mode, the clothing recommendation unit 130 may match (134) the clothing information output from the stylist module 120 with the clothing owned by the user and recommend (135) matching clothing.

In the product recommendation mode, the clothing recommendation unit 130 matches the clothing information output from the stylist module 120 with the clothing information retrieved from the clothing product DB 138 to match recommended products (136) and recommends matching products (136). 137) can. Here, the clothing product DB 138 may be included in the server 200 .

As described above, the embodiment has an effect of recommending clothes considering the user's preference by using the stylist module, which is a deep learning model.

7 is a flowchart illustrating a method for recommending clothes according to an embodiment.

Referring to FIG. 7 , the clothing recommendation method according to the embodiment includes collecting clothing vector information (S100), configuring the clothing vector information and pre-collected style vector information as coordinating vectors, and using the coordinating vectors as inputs. training the stylist module (S200), inputting any one or more of the user's clothing information and the user's selected style information to the stylist module, and based on the clothing information output from the stylist module, the user It may include a step of recommending clothing to the person (S300). Here, the clothing recommendation method may be performed in a clothing recommendation device.

In step S100, the clothing vector collecting unit may collect clothing information and generate a clothing encoding vector in the form of a vector.

In step S200, the stylist module may generate a coordinating vector using a clothing encoding vector and a pre-collected style vector.

In step S200, the stylist module may be trained using the coordinating vectors. The stylist module can be trained using coordinating vectors in which clothing information is not lost. The starry list module can be trained using coordinating vectors in which clothing information is partially lost.

In step S300, the clothing recommendation unit may recommend clothing using clothing information or style information input from the user.

In step S300, the clothing recommendation unit may enter a coordinating mode or a product recommendation mode by user selection.

Various embodiments of the present document are software (eg, machine-readable storage media) (eg, memory (internal memory or external memory)) including instructions stored in a storage medium readable by a machine (eg, a computer). : program). A device is a device capable of calling a stored command from a storage medium and operating according to the called command, and may include an electronic device according to the disclosed embodiments. When the command is executed by the control unit, the control unit may perform a function corresponding to the command directly or by using other components under the control of the control unit. An instruction may include code generated or executed by a compiler or interpreter. The device-readable storage medium may be provided in the form of a non-transitory storage medium. Here, non-temporary means that the storage medium does not contain a signal and is tangible, but does not distinguish whether data is stored semi-permanently or temporarily in the storage medium.

According to an embodiment, a method according to various embodiments disclosed in this document may be included in a computer program product and provided.

According to an embodiment, a computer readable recording medium storing a computer program is provided, comprising the steps of collecting clothing vector information, constructing the clothing vector information and pre-collected style vector information into coordinating vectors, and converting the coordinating vectors into coordinating vectors. training a stylist module as an input, inputting any one or more of clothing information of a user and style information selected by the user to the stylist module, and providing information to the user based on the clothing information output from the stylist module It may include instructions for causing a processor to perform a method including an operation for performing the step of recommending clothes.

According to an embodiment, a computer program stored in a computer-readable recording medium includes collecting clothing vector information, constructing the clothing vector information and pre-collected style vector information into a coordinating vector, and generating the coordinating vector. training a stylist module as an input, inputting any one or more of clothing information of a user and style information selected by the user to the stylist module, and providing information to the user based on the clothing information output from the stylist module It may include instructions for causing a processor to perform a method including an operation for performing the step of recommending clothes.

Although the above has been described with reference to drawings and embodiments, it is understood that those skilled in the art can modify and change the embodiments in various ways without departing from the technical spirit of the embodiments described in the claims below. You will be able to.

100: clothing recommendation device
110: clothing vector collection unit
120: stylist module
130: clothing recommendation unit
200: server
300: user terminal

Claims (16)

a clothing vector collecting unit that collects clothing vector information;
a stylist module that configures the clothing vector information and pre-collected style vector information as coordinating vectors and is trained using the coordinating vectors as inputs; and
a clothing recommendation unit inputting any one or more of clothing information of a user and style information selected by the user to the stylist module, and recommending clothing to the user based on the clothing information output from the stylist module;
Clothing recommendation device comprising a. According to claim 1,
The clothing vector collection unit,
A clothing information collecting unit extracting clothing information from pre-collected clothing images using an artificial neural network;
and a clothing information encoding unit configured to convert the extracted clothing information into the clothing vector information in a vector form. According to claim 1,
The clothing vector collection unit,
a clothing information collection unit that extracts clothing information by matching pre-collected product tags and product information with a clothing information database;
and a clothing information encoding unit configured to convert the extracted clothing information into the clothing vector information in a vector form. According to claim 1,
The stylist module,
Among the clothing vector information and pre-collected style vector information, vector information in which information is not lost is configured as a coordinating vector for primary training;
An apparatus for recommending clothing for secondary training by constructing coordinating vector information from among the clothing vector information and pre-collected style vector information, in which some information is lost. According to claim 1,
The clothing recommendation unit,
An apparatus for recommending clothes to enter a coordinating recommendation mode or a product recommendation mode based on the input information selected by the user. According to claim 5,
In the coordination recommendation mode,
The clothing recommendation unit recommends clothing by matching the clothing information output from the stylist module with clothing information possessed by the user. According to claim 5,
In the product recommendation mode,
The clothing recommendation unit recommends clothing by matching the clothing information output from the stylist module with clothing information retrieved from an external clothing product DB. collecting clothing vector information;
configuring the clothing vector information and pre-collected style vector information into coordinating vectors, and training a stylist module using the coordinating vectors as inputs; and
inputting any one or more of clothing information of a user and style information selected by the user to the stylist module, and recommending clothing to the user based on the clothing information output from the stylist module;
Apparel recommendation method comprising a. According to claim 8,
The step of collecting the clothing vector information,
Extracting clothing information from pre-collected clothing images using an artificial neural network;
and converting the extracted clothing information into the clothing vector information in a vector form. According to claim 8,
The step of collecting the clothing vector information,
extracting clothing information by matching previously collected product tags and product information with a clothing information database;
and converting the extracted clothing information into the clothing vector information in a vector form. According to claim 8,
Training the stylist module,
Among the clothing vector information and pre-collected style vector information, vector information with no information loss among the clothing vector information and pre-collected style vector information is configured as coordinating vectors for primary training, and vector information in which some information among the clothing vector information and pre-collected style vector information is lost A clothing recommendation method configured with coordinating vectors for secondary training. According to claim 8,
The step of recommending the clothing,
A clothing recommendation method in which the user enters a coordinating recommendation mode or a product recommendation mode based on the selected input information. According to claim 12,
In the coordination recommendation mode,
The clothing recommendation step recommends clothing by matching the clothing information output from the stylist module with clothing information owned by the user. According to claim 12,
In the product recommendation mode,
The clothing recommendation step is a method of recommending clothing by matching clothing information output from the stylist module with clothing information retrieved from an external clothing product DB. A computer-readable recording medium storing a computer program,
When the computer program is executed by a processor,
collecting clothing vector information;
configuring the clothing vector information and pre-collected style vector information into coordinating vectors, and training a stylist module using the coordinating vectors as inputs; and
inputting any one or more of clothing information of a user and style information selected by the user to the stylist module, and recommending clothing to the user based on the clothing information output from the stylist module;
A computer-readable recording medium containing instructions for causing the processor to perform an operation including a. As a computer program stored on a computer-readable recording medium,
When the computer program is executed by a processor,
collecting clothing vector information;
configuring the clothing vector information and pre-collected style vector information into coordinating vectors, and training a stylist module using the coordinating vectors as inputs; and
inputting any one or more of clothing information of a user and style information selected by the user to the stylist module, and recommending clothing to the user based on the clothing information output from the stylist module;
A computer program comprising instructions for causing the processor to perform an operation including a.

Images