CN115905593A - Method and system for recommending existing clothes to be worn and put on based on current season style - Google Patents

Method and system for recommending existing clothes to be worn and put on based on current season style Download PDF

Info

Publication number
CN115905593A
CN115905593A CN202211326273.6A CN202211326273A CN115905593A CN 115905593 A CN115905593 A CN 115905593A CN 202211326273 A CN202211326273 A CN 202211326273A CN 115905593 A CN115905593 A CN 115905593A
Authority
CN
China
Prior art keywords
data
clothes
clothing
image
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202211326273.6A
Other languages
Chinese (zh)
Other versions
CN115905593B (en
Inventor
张婷
吴怡桢
柴春雷
周磊晶
王冠云
陶冶
张东亮
赵晓亮
宋兴辉
高子惠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN202211326273.6A priority Critical patent/CN115905593B/en
Publication of CN115905593A publication Critical patent/CN115905593A/en
Application granted granted Critical
Publication of CN115905593B publication Critical patent/CN115905593B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02PCLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30Computing systems specially adapted for manufacturing

Landscapes

  • Image Analysis (AREA)

Abstract

The invention belongs to the technical field of clothing recommendation, and provides a method and a system for wearing and putting on existing clothing for recommendation based on the current trend style, wherein a product database of current trend clothing is established through a terminal, personal data of a user are obtained, similarity calculation is carried out on image characteristics and text characteristics extracted from the personal data in the product database, target clothing is obtained by matching, then matching and sequencing are carried out on the target clothing, the maximum possibility of matching the existing clothing is realized to follow the current trend, intelligent clothing matching is realized, and the utilization rate of the existing clothing is improved; the invention can ensure the comprehensive analysis of the characteristics of the fashion by extracting the image characteristics and the text characteristics from the product database and the personal data of the fashion clothing in season, and simultaneously, the user inputs the existing clothing and the image data in an image and character mode, so that the characteristics of the user can not be lost when pursuing the fashion, thereby realizing the aim of pursuing the fashion without blindness.

Description

Method and system for recommending existing clothes to be worn and put on based on current season style
Technical Field
The invention belongs to the technical field of clothing recommendation, and particularly relates to a method and a system for recommending existing clothing wearing based on the current season style.
Background
In daily life, along with the improvement of living standard, people pay more and more attention to the management of personal images, and every day, people ask what to wear by themselves. The wardrobe may contain various clothes and continuously purchases new clothes, but with the continuous change of trend, the clothes in the wardrobe are more and more, and the idle clothes are also more. It is a challenge to determine what clothing combination the most probable collocation follows the current season and maximize our visual appeal.
In the prior art, a collocation template is also provided, and although the problems of partial collocation and recommendation can be solved, the specialty and the user pertinence are obviously not strong enough; a great gap is left to be filled in for the research on the existing clothing putting-on recommendation based on the current season style.
Disclosure of Invention
The invention provides a method and a system for recommending the wearing of existing clothes based on the current trend style of the season, which match target clothes for a user by acquiring the image and text information of the current trend style of the season and the image and text information of the existing clothes and personal images and performing similarity calculation through personal data and a product database of the current trend clothes, match the target clothes with the user, and can follow the current trend by using the existing clothes matching by matching the target clothes, thereby improving the utilization rate of the existing clothes and intelligent clothes matching.
In order to achieve the above object, the present invention provides the following technical solutions:
the invention provides a method for recommending existing clothes to be worn based on the current season style, which comprises the following steps:
(1) The terminal establishes a product database of the current tide clothes; the product database comprises a current trend decoration image database and a current trend character library in the season;
(2) The terminal acquires personal data of a user; the personal data comprises existing clothes data and personal image data;
(3) The terminal carries out similarity calculation in a product database according to the image features and the text features extracted from the personal data, matches target clothes for the user and sorts the clothes;
(4) Matching the target clothes matched by the user according to matching rules to generate a plurality of combined recommended data clothes wearing and matching;
(5) And the terminal analyzes and stores the collocation preference finally selected by the user and sets the collocation preference as a preferred collocation rule of subsequent collocation.
The matching results are sorted according to cosine similarity, the matching results are most similar to current fashion clothes in priority, but a user may select the first matching according to subjective selection of the user instead of selecting the first matching, the matching rule is matched according to a plurality of rules of color matching rules, category matching rules and occasion matching rules, the emphasis of each displayed finally matched clothes on color, category or occasion is different, the terminal analyzes which rule of the current more important color, category or occasion of the user according to the final selection of the user, and the analysis result is stored as the matching rule which is preferably considered for next matching, so that the whole matching is more intelligent.
The terminal can directly use the display screen to display the clothing information, and the user selects the matching result through the display screen, so that the user can more intuitively know the matching result; the terminal may further send the clothing information corresponding to the clothing to a portable terminal corresponding to the user, such as a mobile phone, a tablet computer, and the like of the user, which is not limited herein.
The terminal may obtain the personal data of the user through its own data collector, or may receive the personal data of the user sent by other terminals, which is not limited herein.
Further, in step (1), the step of establishing the product database comprises:
the terminal collects an image database and a text related to the current fashion clothing through crawler software, preprocesses the image database and the text, extracts features one by one, and finally stores the feature data to establish a product database;
the characteristic data comprises color data, style data, size data, collocation data, material data, element data, occasion data and text description of the current fashion clothing as corresponding appearance adjectives.
Further, in the step (2), the process of acquiring the personal data is as follows:
the terminal obtains the existing clothes data and the personal image data through the character description and the offline shooting of the user, preprocesses the existing clothes data and extracts the features of the existing clothes data and the personal image data one by one;
the character data includes gender, age, height and dimension, and specifically, the personal character data includes a picture of the user himself and a text description which can be described by the gender, age, height and dimension. The existing clothes data comprise color data, style data, size data, matching data, material data, element data and occasion data, and specifically, the existing clothes data comprise text descriptions of colors, styles, sizes, matches, materials, elements and occasions of the existing clothes by taking pictures of the existing clothes off line; further, the process of preprocessing the product database or the personal data and extracting the features one by one includes:
the data preprocessing comprises classification and screening of input images, resetting of sizes and cleaning of text information; preferably, the text information includes trending words and word descriptions input by the user. Performing auxiliary classification on input image data according to human body parts by adopting a trained image classification deep learning model;
scaling the image data image size to w x h;
extracting keywords of the text information;
carrying out independent convolution processing on the collected tidal current clothing image or the existing clothing data, and extracting semantic features of corresponding texts;
the convolution processing comprises feature extraction and feature mapping, and the feature extraction operation process is as follows:
Figure BDA0003912213860000041
wherein,
Figure BDA0003912213860000045
the j-th operation result of the I layer is represented;
f () represents an activation function;
Figure BDA0003912213860000046
an index set representing a plurality of input feature maps corresponding to the jth output feature map of the nth layer;
Figure BDA0003912213860000047
representing a bias term, common to all input feature maps;
Figure BDA0003912213860000048
a convolution kernel representing the size of one I x j in layer I;
the calculation process of the feature mapping layer is as follows:
Figure BDA0003912213860000042
wherein down () represents a down-sampling function that sums all pixels in different blocks of the input image, thereby enabling the output image to be reduced to the original in two dimensions
Figure BDA0003912213860000043
f () represents an activation function;
Figure BDA0003912213860000044
is a multiplicative paranoia term;
the convolution processing is a feature mapping F learned by a deep learning model, and the calculation formula is as follows:
v=F(image);
wherein v represents a feature vector;
image is an image of size 3 × w × h;
3 represents RGB; v is 1 x 128 in size.
The image feature network can be a network such as VGG, resNet, inclusion and the like.
The semantic feature extraction of the text comprises the description of the style, style and color of the garment, and the keyword feature extraction model can be an NLP training model such as Bert, transform-XL or ULMFiT.
Storing all image data and text data in a product database or personal data in a trained image classification deep learning model respectively to obtain a trend clothing feature vector v s Or personal data feature vector v t (ii) a The trend clothes feature vector v s Image feature vectors and text for trending clothingAdding the feature vectors; person data feature vector v t The image feature vector and the text feature vector for the personal data are added.
Further, the matching rule in the step (3) is as follows:
taking the personal image data of the user as a reference, carrying out similarity calculation on the existing clothing data and a product database of the current trend clothing, and matching target clothing for the user;
determining the existing clothes corresponding to the matched tidal current clothes data as target clothes matched with the user;
the target garment includes an item of apparel and/or a combination of items of apparel.
Further, the similarity is calculated by using a cosine similarity function, and the calculation formula is as follows:
in order to realize the purpose of the method,
Figure BDA0003912213860000051
wherein similarity () is a cosine similarity function;
n is the number of samples;
v s is a trend clothes feature vector;
v t is a personal data feature vector.
The similarity calculation refers to that the existing clothing data of the user is put forward a corresponding trend clothing feature vector v through a product database of the current trend clothing in the season t And comparing the vectors according to the feature vectors of the related types stored in the product database of the current fashion clothing in the season, and calculating the similarity between the vectors and the feature vectors.
Further, the target clothes ordering rule in the step (4) is as follows:
calculating and comparing cosine similarity of the existing clothes, and taking similarity (v) t ,v s ) The value of the similarity score is closest to 1, the Z pieces of clothes with the similarity scores closest to 1 are recommended, the Z pieces of clothes are sorted according to the sequence of the similarity scores from large to small, and the proper matching is recommended according to the personalized demand information of the user. Further, in the step (4), matching with a gaugeThe method comprises one or more of color matching rules, category matching rules, occasion matching rules, user preferences, clothes storage time, scene data and season data, and the clothes matching rules are determined according to the personal data of the user, the product information of the current fashion clothes, the user preferences, the scene data and the season data. Specifically, the combination between the two most pleasing colors can be calculated by adding or subtracting the isomorphic color numbers of the color matching rules according to the matching modes of the color matching rules, the category matching rules and the occasion matching rules, namely the matching modes of internal and external lapping, up and down lapping and three pieces of matching; the collocation rules are set according to colors, categories and occasions, the prior art has a plurality of practical rules for clothing collocation, and the collocation rules in the invention do not belong to protection contents, so the detailed description is not provided.
Preferably, the terminal is a server terminal; the intelligent clothing matching system is used for storing and processing a large amount of data, and can perform intelligent clothing matching service for a large number of users; the terminal can also be a portable terminal, such as a mobile phone, a tablet computer, a notebook, a personal computer and other terminals, and can provide portable intelligent clothing matching service for users of the terminal.
The invention also provides a system for recommending the existing clothing wearing based on the current season trend style, which comprises:
the first obtaining unit is used for establishing a product database of the current trend clothes, the product database comprises a current trend clothes image database and a trend text database, the product database comprises images of the current trend clothes and corresponding text descriptions thereof, specifically comprises color data, style data, size data, collocation data, material data, element data and occasion data, each image is subjected to feature extraction through the first image feature extraction unit, feature vectors of m, n, h are extracted, keyword features are extracted from the corresponding text descriptions through the first text feature extraction unit, and feature vectors of m, n, h are extracted as well; adding the image of the fashion clothing and the feature vector of the text, and performing deeper feature extraction on the image through a deep convolutional layer;
a second acquisition unit for acquiring personal data of a user, the personal data including image data and existing clothing data; the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring existing clothing data of a user, and the existing clothing data comprises color data, style data, preference data, time data and occasion data; extracting features of each image through a second image feature extraction unit to extract feature vectors of m x n x h, extracting keyword features of corresponding text descriptions through a second text feature extraction unit, and extracting the feature vectors of m x n x h in the same way; adding the image of the personal data and the feature vector of the text, and extracting the features of the text in a deeper layer through a deep convolutional layer;
the terminal extracts image characteristics and text characteristics according to the personal data and carries out similarity calculation in a product database to match target clothes for the user, wherein the target clothes comprise clothes items and/or combinations of the clothes items;
a collocation unit: matching target clothes matched by the user according to matching rules to generate a plurality of combined recommended data clothes wearing and matching, wherein the matching rules comprise user preference, clothes warehousing time, scene data and season data;
a sorting unit: sorting the matching units according to the similarity scores and the matching rules;
a storage unit: and analyzing and storing the collocation preference finally selected by the user, and setting the collocation preference as a preferred rule of the subsequent collocation.
Further, the sorting rule calculates and compares by cosine similarity, and takes similarity (v) t ,v s ) The value of the similarity score is closest to 1, the Z pieces of clothing with the similarity score closest to 1 are recommended, the Z pieces of clothing are sorted according to the sequence of the similarity scores from large to small, and the proper matching is recommended according to the personalized demand information of the user.
The invention has the following beneficial effects:
(1) The method comprises the steps of establishing a product database of the current season trend clothes through a terminal, obtaining personal data of a user, carrying out similarity calculation on image features and text features extracted from the personal data in the product database, matching to obtain target clothes, and then carrying out matching and sequencing on the target clothes, so that the maximum possibility matching of the existing clothes is realized to follow the current season trend, intelligent clothes matching is realized, and the utilization rate of the existing clothes is improved;
(2) The invention can ensure that the characteristics of the fashion clothes in season can be comprehensively analyzed by extracting the image characteristics and the text characteristics from the product database and the personal data of the fashion clothes in season, and meanwhile, the user inputs the existing clothes and the image data in an image and character mode, so that the characteristics of the user can not be lost when pursuing the fashion clothes in season, and the aim of pursuing the fashion clothes in no blindness can be realized;
(3) The matched clothes are sorted by adopting the cosine similarity, so that the sorting result is closer to the current style of the season;
(4) The terminal analyzes which rule of the current heavier colors, categories or occasions of the user is selected according to the final selection of the user, and stores the analysis result as a matching rule which is preferably considered for next matching, so that more intelligent clothes matching is realized;
(5) The method and the system can automatically generate the recommended clothing matching according to the requirements input by the user, and realize intelligent clothing matching to meet the wearing and putting requirements of the user.
Drawings
FIG. 1 is a flow chart of a method of punch-through recommendation of the present invention.
FIG. 2 is a flow chart of the present invention for building a product database.
Fig. 3 is a flow chart of personal data acquisition of the present invention.
FIG. 4 is a matching rule flow diagram of the present invention.
Fig. 5 is a schematic diagram of data processing of the first acquisition unit and the second acquisition unit of the present invention.
Fig. 6 is a schematic diagram of the system of the present invention.
Detailed Description
The following detailed description of the embodiments of the present invention is provided in conjunction with the accompanying drawings, and it should be noted that the embodiments are merely illustrative of the present invention and should not be considered as limiting the invention, and the purpose of the embodiments is to make those skilled in the art better understand and reproduce the technical solutions of the present invention, and the protection scope of the present invention should be subject to the scope defined by the claims.
As shown in fig. 1, the invention provides a method for recommending existing clothing wearing based on the current season fashion style, which comprises the following steps:
s1, a terminal establishes a product database of the current tide costume; the product database comprises a current trend decoration image database and a current trend character library in the season;
as shown in fig. 2, the step of building a product database includes:
s11, acquiring an image database and a text related to the current fashion clothing by the terminal through crawler software;
s12, preprocessing the image and extracting features of the image one by one;
s13, storing the characteristic data to establish a product database;
the characteristic data comprises color data, style data, size data, collocation data, material data, element data, occasion data and text description of the current fashion clothing as corresponding appearance adjectives.
The process of preprocessing the product database and extracting the features one by one is as follows:
the data preprocessing comprises classification and screening of input images, resetting of sizes and cleaning of text information; preferably, the text information includes trending words and word descriptions input by the user. Performing auxiliary classification on input image data according to human body parts by adopting a trained image classification deep learning model;
scaling the image data image size to w x h;
extracting keywords of the text information;
carrying out independent convolution processing on the collected tidal current clothing image, and extracting semantic features of corresponding texts;
s2, the terminal acquires personal data of a user; the personal data comprises existing clothing data and personal image data;
as shown in fig. 3, the process of personal data acquisition is:
s21, the terminal obtains the existing clothing data and the personal image data through the character description and the offline shooting of the user;
s22, preprocessing the image and extracting the features of the image one by one;
the character data includes gender, age, height and dimension, and specifically, the personal character data includes a picture of the user himself and a text description which can be described by the gender, age, height and dimension. The existing clothing data comprises color data, style data, size data, matching data, material data, element data and occasion data, and specifically, the existing clothing data comprises text descriptions of colors, styles, sizes, matches, materials, elements and occasions of the existing clothing through offline photographing of the existing clothing.
The process of preprocessing the personal data and extracting the features one by one is as follows:
the data preprocessing comprises classification and screening of input images, resetting of sizes and cleaning of text information; preferably, the text information includes trending words and word descriptions input by the user. Performing auxiliary classification on input image data according to human body parts by adopting a trained image classification deep learning model;
scaling the image data image size to w x h;
extracting keywords of the text information;
carrying out independent convolution processing on the acquired existing clothing data, and extracting semantic features of corresponding texts;
the convolution processing comprises feature extraction and feature mapping, and the feature extraction and operation process is as follows:
Figure BDA0003912213860000111
wherein,
Figure BDA0003912213860000115
the j-th operation result of the I layer is represented;
f () represents an activation function;
Figure BDA0003912213860000116
an index set representing a plurality of input feature maps corresponding to the jth output feature map of the nth layer;
Figure BDA0003912213860000117
representing a bias term, common to all input profiles;
Figure BDA0003912213860000118
a convolution kernel representing the size of one I x j in layer I;
the calculation process of the feature mapping layer is as follows:
Figure BDA0003912213860000112
wherein down () represents a down-sampling function that sums all pixels in different blocks of the input image, thereby enabling the output image to be reduced to the original in two dimensions
Figure BDA0003912213860000113
f () represents an activation function;
Figure BDA0003912213860000114
is a multiplicative bias term;
the convolution processing is a feature mapping F learned by a deep learning model, and the calculation formula is as follows:
v=F(image);
wherein v represents a feature vector;
image is an image of size 3 × w × h;
3 denotes RGB, v being 1 × 128 in size.
The deep learning model selected for the convolution processing can be VGG, resNet and inclusion.
The semantic feature extraction of the text comprises the description of the style, style and color of the garment, and the keyword feature extraction model can be an NLP training model such as Bert, transform-XL or ULMFiT.
Storing all image data and text data in a product database or a personal database in a trained image classification deep learning model respectively to obtain a trend clothing feature vector v s Or personal data feature vector v t (ii) a The trend clothing feature vector v s Adding the image characteristic vector and the text characteristic vector of the fashion clothing; person data feature vector v t The image feature vector and the text feature vector for the personal data are added.
S3, the terminal carries out similarity calculation in a product database according to the image characteristics and the text characteristics extracted from the personal data, matches target clothes for the user and sorts the target clothes;
as shown in fig. 4, the matching rule is:
s31, taking the personal image data of the user as a reference, carrying out similarity calculation on the existing clothing data and a product database of the fashion clothing in the season, and matching target clothing for the user;
s32, determining the existing clothes corresponding to the matched tide clothes data as target clothes matched with the user;
the target garment includes an item of apparel and/or a combination of items of apparel.
S4, matching the target clothes matched by the user according to matching rules to generate a plurality of combined recommended data clothes matching;
the similarity is calculated by adopting a cosine similarity function, and the calculation formula is as follows:
in order to realize the purpose of the method,
Figure BDA0003912213860000131
wherein similarity () is a cosine similarity function;
n is the number of samples;
v s is a trend clothes feature vector;
v t is a personal data feature vector.
Calculating and comparing cosine similarity of the existing clothes, and taking similarity (v) t ,v s ) The value of the similarity score is closest to 1, the Z pieces of clothing with the similarity score closest to 1 are recommended, the Z pieces of clothing are sorted according to the sequence of the similarity scores from large to small, and the proper matching is recommended according to the personalized demand information of the user. The similarity calculation refers to that the existing clothing data of the user is put forward a corresponding trend clothing feature vector v through a product database of the current trend clothing in the season t And comparing the vectors according to the related types of characteristic vectors stored in the product database of the fashion clothing in the season, and calculating the similarity between the vectors and the characteristic vectors. The matching rules comprise one or more of color matching rules, category matching rules, occasion matching rules, user preferences, clothes storage time, scene data and season data, and the clothes matching rules are determined according to the personal data of the user, the product information of the current fashion clothes, the user preferences, the scene data and the season data. Specifically, the combination between the two most pleasing colors can be calculated by adding or subtracting the isomorphic color numbers of the color matching rules according to the matching modes of the color matching rules, the category matching rules and the occasion matching rules, namely the matching modes of internal and external lapping, up and down lapping and three pieces of matching; the collocation rules are set according to colors, categories and occasions, the prior art has a plurality of practical rules for clothing collocation, and the collocation rules in the invention do not belong to protection contents, so the detailed description is not provided.
And S5, the terminal analyzes and stores the collocation preference finally selected by the user, and sets the collocation preference as a preferred collocation rule of subsequent collocation.
The matching results are sorted according to the cosine similarity, the matching result is preferably similar to current fashion clothes, but a user may select the first matching according to subjective selection of the user rather than selecting the first matching, the matching rule is matched according to a plurality of rules of color matching rules, category matching rules and occasion matching rules, the emphasis of each displayed finally-matched clothes on color, category or occasion is different, the terminal analyzes which rule of the user which is more important in color, category or occasion at present according to the final selection of the user, and the analysis result is stored as the matching rule which is preferably considered in next matching, so that the whole matching is more intelligent.
The terminal can directly use the display screen to display the clothing information, and the user selects the matching result through the display screen, so that the user can more intuitively know the matching result; the terminal may further send the clothing information corresponding to the clothing to a portable terminal corresponding to the user, such as a mobile phone, a tablet computer, and the like of the user, which is not limited herein.
The terminal may obtain the personal data of the user through its own data collector, or may receive the personal data of the user sent by other terminals, which is not limited herein.
Preferably, the terminal is a server terminal; the intelligent clothing matching system is used for storing and processing a large amount of data, and can perform intelligent clothing matching service for a large number of users; the terminal can also be a portable terminal, such as a mobile phone, a tablet computer, a notebook, a personal computer and other terminals, and can provide portable intelligent clothing matching service for users of the terminal.
As shown in fig. 5 to 6, the present invention further provides a system for recommending existing clothing wearing based on the current season trend style, the system comprising:
the first obtaining unit 1 is used for establishing a product database of the current trend clothes, the product database comprises a current trend clothes image database and a trend text database, the product database comprises an image of the current trend clothes and corresponding text descriptions thereof, and specifically comprises color data, style data, size data, collocation data, material data, element data and occasion data, each image is subjected to feature extraction through a first image feature extraction unit, feature vectors of m & ltn & gt & lth & gt are extracted, keyword features are extracted from the corresponding text descriptions through a first text feature extraction unit, and feature vectors of m & ltn & gt & lth & gt are extracted as well; adding the image of the fashion clothing and the feature vector of the text, and performing deeper feature extraction on the image through a deep convolutional layer;
a second obtaining unit 2 for obtaining personal data of a user, the personal data including figure data and existing clothes data; the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring existing clothing data of a user, and the existing clothing data comprises color data, style data, preference data, time data and occasion data; extracting features of each image through a second image feature extraction unit to extract feature vectors of m x n x h, extracting keyword features of corresponding text descriptions through a second text feature extraction unit, and extracting the feature vectors of m x n x h in the same way; adding the image of the personal data and the feature vector of the text, and extracting the features of the text in a deeper layer through a deep convolutional layer;
preferably, the first picture feature extraction unit, the first text feature extraction unit, the second picture feature extraction unit and the second text feature extraction unit may extract using a convolution process and a keyword feature extraction model.
The matching unit 3 is used for performing similarity calculation in a product database by the terminal according to the image characteristics and the text characteristics extracted by the personal data to match target clothes for the user, wherein the target clothes comprise clothes single items and/or combinations of the clothes single items;
and a collocation unit 4: matching target clothes matched by the user according to matching rules to generate a plurality of combined recommended data clothes wearing and matching, wherein the matching rules comprise user preference, clothes warehousing time, scene data and season data;
the sorting unit 5: sorting the matching unit matches according to the similarity scores;
the storage unit 6: and analyzing and storing the collocation preference finally selected by the user, and setting the collocation preference as a preferred rule of the subsequent collocation.
In some preferred schemes, the sorting rule calculates and compares the similarity (v) by cosine similarity, and takes similarity (v) t ,v s ) The value of the similarity score is closest to 1, the Z pieces of clothing with the similarity score closest to 1 are recommended, the Z pieces of clothing are sorted according to the sequence of the similarity scores from large to small, and the proper matching is recommended according to the personalized demand information of the user.
The method comprises the steps of establishing a product database of the current season trend clothes through a terminal, obtaining personal data of a user, carrying out similarity calculation on image features and text features extracted from the personal data in the product database, matching to obtain target clothes, and then carrying out matching and sequencing on the target clothes, so that the maximum possibility matching of the existing clothes is realized to follow the current season trend, intelligent clothes matching is realized, and the utilization rate of the existing clothes is improved;
the invention can ensure that the characteristics of the fashion clothes in season can be comprehensively analyzed by extracting the image characteristics and the text characteristics from the product database and the personal data of the fashion clothes in season, and simultaneously, the user inputs the existing clothes and the image data in an image and character mode, so that the characteristics of the user can not be lost when pursuing the fashion clothes in season, and the aim of pursuing the fashion without blindness is fulfilled. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the present application.

Claims (10)

1. A method for recommending existing clothes wearing based on current season style is characterized by comprising the following steps:
(1) The terminal establishes a product database of the current tide clothes; the product database comprises a current trend decoration image database and a current trend character library in the season;
(2) The terminal acquires personal data of a user; the personal data comprises existing clothes data and personal image data;
(3) The terminal carries out similarity calculation in a product database according to the image features and the text features extracted from the personal data, matches target clothes for the user and sorts the clothes;
(4) Matching the target clothes matched by the user according to matching rules to generate a plurality of combined recommended data clothes wearing and matching;
(5) And the terminal analyzes and stores the collocation preference finally selected by the user and sets the collocation preference as a preferred collocation rule of subsequent collocation.
2. The method for recommending existing clothes wearing and putting on based on the current season trend style according to claim 1, wherein in the step (1), the step of establishing the product database comprises:
the terminal collects an image database and a text related to the current tidal current clothing through crawler software, preprocesses the image database and the text, extracts the characteristics of the images one by one, and finally stores the characteristic data of the tidal current clothing to establish a product database;
the tidal current clothing feature data comprise color data, style data, size data, collocation data, material data, element data, occasion data and text description of the current tidal current clothing as corresponding appearance adjectives.
3. The method for recommending existing clothes wearing based on the current season trend style according to claim 2, wherein in the step (2), the personal data acquisition process comprises:
the terminal obtains the existing clothes data and the personal image data through the character description and the offline shooting of the user, preprocesses the existing clothes data and extracts the features of the existing clothes data and the personal image data one by one;
the types are distinguished according to the wearing position of the clothes, and comprise a hat, a scarf, a coat, a lower garment and a one-piece garment; the image data comprises gender, age, height and dimensions;
the existing clothing data comprises color data, style data, size data, collocation data, material data, element data and occasion data.
4. The method for recommending existing clothes wearing and putting on based on the current season trend style as claimed in claim 3, wherein the pre-processing and class-by-class feature extraction processes of the product database or the personal data are as follows: performing auxiliary classification on input image data according to the wearing position of the clothes by adopting a trained image classification deep learning model;
scaling the image data image size to w x h;
extracting keywords of the text information;
carrying out independent convolution processing on the collected tidal current clothing image or the existing clothing data, and extracting semantic features of corresponding texts;
storing all image data and text data in a product database or personal data in a trained image classification deep learning model respectively to obtain a trend clothing feature vector v s Or personal data feature vector v t
The trend clothing feature vector v s Adding the image characteristic vector and the text characteristic vector of the fashion clothing;
personal data feature vector v t The image feature vector and the text feature vector for the personal data are added.
5. The method for recommending existing clothes wearing and putting up based on the current season trend style as claimed in claim 3, wherein the matching rule in the step (3) is as follows:
taking the personal image data of the user as a reference, carrying out similarity calculation on the existing clothing data and a product database of the current trend clothing, and matching target clothing for the user;
determining the existing clothes corresponding to the matched tidal current clothes data as target clothes matched with the user;
the target garment includes an item of apparel and/or a combination of items of apparel.
6. The method for recommending existing clothing wearing and putting up based on the current season trend style as claimed in claim 5, wherein the similarity is calculated by using a cosine similarity function, and the calculation formula is as follows:
Figure FDA0003912213850000031
wherein similarity () is a cosine similarity function;
n is the number of samples;
v s is a trend clothing feature vector;
v t is a personal data feature vector.
7. The method for recommending the wearing of existing clothes based on the current season trend style as claimed in claim 6, wherein the method for ordering the target clothes is as follows: taking personal image data of the user as a reference, calculating and comparing cosine similarity of the existing clothes and the tidal current clothes, and taking similarity (v) t ,v s ) And recommending the Z-piece garment with the similarity score closest to 1, and sorting the Z-piece garments with the similarity scores closest to 1 from big to small.
8. The method according to claim 1, wherein in the step (4), the matching rules include one or more of color matching rules, category matching rules, occasion matching rules, user preferences, time for putting clothes in storage, scene data and season data, and the clothes matching rules are determined according to the personal data of the user, the product information of the current fashion clothes, the user preferences and the scene data.
9. A system for recommending existing clothing wearing based on current season trend style, characterized in that the system is adapted to the method of any one of claims 1-8: the system comprises:
the first obtaining unit is used for establishing a product database of the current trend clothes, the product database comprises images of the current trend clothes and corresponding text descriptions, specifically comprises color data, style data, size data, collocation data, material data, element data and occasion data, each image is subjected to feature extraction through the first image feature extraction unit, feature vectors of m & ltn & gt & lth & gt are extracted, keyword features are extracted from the corresponding text descriptions through the first text feature extraction unit, and the feature vectors of m & ltn & gt & lth & gt are extracted in the same way; adding the image of the fashion clothing and the feature vector of the text, and performing deeper feature extraction on the image through a deep convolution layer;
a second acquisition unit for acquiring personal data of a user, the personal data including image data and existing clothing data; the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring existing clothing data of a user, and the existing clothing data comprises color data, style data, preference data, time data and occasion data; extracting features of each image through a second image feature extraction unit to extract feature vectors of m x n x h, extracting keyword features of corresponding text descriptions through a second text feature extraction unit, and extracting the feature vectors of m x n x h in the same way; adding the image of the personal data and the feature vector of the text, and extracting the features of the text in a deeper layer through a deep convolutional layer;
the terminal extracts image characteristics and text characteristics according to the personal data and carries out similarity calculation in a product database to match target clothes for the user, wherein the target clothes comprise clothes items and/or combinations of the clothes items;
a collocation unit: matching target clothes matched by the user according to matching rules to generate a plurality of combined recommended data clothes wearing and matching, wherein the matching rules comprise user preference, clothes warehousing time, scene data and season data;
a sorting unit: sorting the matching unit matches according to the similarity scores and the individual requirements of the users;
a storage unit: and analyzing and storing the collocation preference finally selected by the user, and setting the collocation preference as a preferred rule of the subsequent collocation.
10. The system for recommending existing clothes for putting on based on current season trend style according to claim 8, wherein the sorting rule is used for calculating and comparing cosine similarity to obtain similarity (v) t ,v s ) The value of the similarity score is closest to 1, the Z pieces of clothing with the similarity score closest to 1 are recommended, the Z pieces of clothing are sorted according to the sequence of the similarity scores from large to small, and the proper matching is recommended according to the personalized demand information of the user.
CN202211326273.6A 2022-10-27 2022-10-27 Method and system for recommending wearing and taking of existing clothing based on current style in season Active CN115905593B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211326273.6A CN115905593B (en) 2022-10-27 2022-10-27 Method and system for recommending wearing and taking of existing clothing based on current style in season

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211326273.6A CN115905593B (en) 2022-10-27 2022-10-27 Method and system for recommending wearing and taking of existing clothing based on current style in season

Publications (2)

Publication Number Publication Date
CN115905593A true CN115905593A (en) 2023-04-04
CN115905593B CN115905593B (en) 2024-07-23

Family

ID=86482638

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211326273.6A Active CN115905593B (en) 2022-10-27 2022-10-27 Method and system for recommending wearing and taking of existing clothing based on current style in season

Country Status (1)

Country Link
CN (1) CN115905593B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116402590A (en) * 2023-06-06 2023-07-07 武汉纺织大学 Personalized clothing mall system based on artificial intelligence and recommendation method
CN116823361A (en) * 2023-08-31 2023-09-29 博洛尼智能科技(青岛)有限公司 Jewelry collocation detection and pushing method based on artificial intelligence

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104331417A (en) * 2014-10-09 2015-02-04 深圳码隆科技有限公司 Matching method for personnel garments of user
CN107679960A (en) * 2017-10-12 2018-02-09 东北大学 A kind of recommendation method of the customized clothing based on image of clothing and label text bimodal content analysis
WO2018228448A1 (en) * 2017-06-16 2018-12-20 深圳市商汤科技有限公司 Method and apparatus for recommending matching clothing, electronic device and storage medium
CN109360050A (en) * 2018-09-27 2019-02-19 东华大学 Personal care garment management and personalized collocation recommendation intelligence system based on perceptual demand
KR20200066970A (en) * 2018-12-03 2020-06-11 손원호 Method and system for automatic, customized outfit coordination, product recommendation and information management based on user's fashion item possession and preference
CN112163930A (en) * 2020-09-27 2021-01-01 深圳莱尔托特科技有限公司 Matching recommendation method and matching recommendation system
US20210182934A1 (en) * 2019-12-13 2021-06-17 The Mada App, LLC System rendering personalized outfit recommendations
CN113034237A (en) * 2020-12-28 2021-06-25 武汉纺织大学 Dress suit recommendation system and method

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104331417A (en) * 2014-10-09 2015-02-04 深圳码隆科技有限公司 Matching method for personnel garments of user
WO2018228448A1 (en) * 2017-06-16 2018-12-20 深圳市商汤科技有限公司 Method and apparatus for recommending matching clothing, electronic device and storage medium
CN107679960A (en) * 2017-10-12 2018-02-09 东北大学 A kind of recommendation method of the customized clothing based on image of clothing and label text bimodal content analysis
CN109360050A (en) * 2018-09-27 2019-02-19 东华大学 Personal care garment management and personalized collocation recommendation intelligence system based on perceptual demand
KR20200066970A (en) * 2018-12-03 2020-06-11 손원호 Method and system for automatic, customized outfit coordination, product recommendation and information management based on user's fashion item possession and preference
US20210182934A1 (en) * 2019-12-13 2021-06-17 The Mada App, LLC System rendering personalized outfit recommendations
CN112163930A (en) * 2020-09-27 2021-01-01 深圳莱尔托特科技有限公司 Matching recommendation method and matching recommendation system
CN113034237A (en) * 2020-12-28 2021-06-25 武汉纺织大学 Dress suit recommendation system and method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
郭青;孙健;: "基于专家系统的个性化推荐系统的设计与实现", 软件工程, no. 06, 5 June 2016 (2016-06-05), pages 26 - 31 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116402590A (en) * 2023-06-06 2023-07-07 武汉纺织大学 Personalized clothing mall system based on artificial intelligence and recommendation method
CN116402590B (en) * 2023-06-06 2023-09-15 武汉纺织大学 Personalized clothing mall system based on artificial intelligence and recommendation method
CN116823361A (en) * 2023-08-31 2023-09-29 博洛尼智能科技(青岛)有限公司 Jewelry collocation detection and pushing method based on artificial intelligence
CN116823361B (en) * 2023-08-31 2023-12-12 博洛尼智能科技(青岛)有限公司 Jewelry collocation detection and pushing method based on artificial intelligence

Also Published As

Publication number Publication date
CN115905593B (en) 2024-07-23

Similar Documents

Publication Publication Date Title
CN105426850B (en) Associated information pushing device and method based on face recognition
US10109051B1 (en) Item recommendation based on feature match
CN115905593B (en) Method and system for recommending wearing and taking of existing clothing based on current style in season
US10963939B1 (en) Computer vision based style profiles
US8897485B2 (en) Determining an interest level for an image
CN112348640B (en) Online shopping system and method based on facial emotion state analysis
US8873851B2 (en) System for presenting high-interest-level images
US9014510B2 (en) Method for presenting high-interest-level images
US20140003737A1 (en) Modifying digital images to increase interest level
US20140002644A1 (en) System for modifying images to increase interestingness
CN108133055A (en) Intelligent dress ornament storage device and based on its storage, recommend method and apparatus
CN116097320A (en) System and method for improved facial attribute classification and use thereof
CN105426462A (en) Image searching method and device based on image element
WO2019011936A1 (en) Method for evaluating an image
CN109597907A (en) Dress ornament management method and device, electronic equipment, storage medium
KR20210002410A (en) System, method and program of constructing dataset for training appearance recognition model
CN115168745B (en) Virtual character image recreating method and system based on image technology
CN107729380A (en) Clothing matching method, terminal, terminal
CN110909746A (en) Clothing recommendation method, related device and equipment
US11544768B2 (en) System and method for fashion recommendations
CN110543813B (en) Face image and gaze counting method and system based on scene
JP2014229129A (en) Combination presentation system and computer program
CN112508114A (en) Intelligent clothing recommendation system and method
Wei et al. Inferring intrinsic correlation between clothing style and wearers’ personality
CN113377970A (en) Information processing method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant