Server, method and computer-readable storage medium for selecting eyewear equipment

Abstract

A server including a processing circuit system, the processing circuit system configured to receive one or more images, the one or more images including one or more representations of people. In addition, the processing circuit system is configured to apply a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, and an aesthetic component score is generated for each of the one or more images. Further, the processing circuit system is configured to generate a user eyewear configuration profile for a user, the user is matched with a persona from a persona database, each persona in the persona database is linked to one or more persona eyewear configuration profiles, the one or more persona eyewear configuration profiles are based on the aesthetic component score, and an eyewear configuration is selected for the user based on the generated user eyewear configuration profile.

Description

Background Art

[0001] The "background technology" description provided herein is for the purpose of generally introducing the context of the present disclosure. To the extent that the work of the presently named inventors is described in this background technology section, and aspects of this description that may not be identified as prior art at the time of filing, are neither explicitly nor implicitly admitted to be prior art against the present invention.

[0002] Online businesses need personalized product recommendations to push the right eyewear to consumers at the right time. For example, some consumers are more sensitive to fashion and trends than any other criteria. Typically, eyewear or product recommendations can be made through morphological rules, analysis of style, lifestyle and / or personality questionnaires, and analysis of consumer browsing history. However, more personalized and accurate eyewear recommendations are needed. Summary of the invention

[0003] According to aspects of the disclosed subject matter, a server includes a processing circuit system configured to receive one or more images, the one or more images including one or more representations of a person. In addition, the processing circuit system is configured to apply a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, and an aesthetic component score is generated for each of the one or more images. Further, the processing circuit system is configured to generate a user eyewear configuration profile for a user, the user is matched with a persona from a persona database, each persona in the persona database is linked to one or more persona eyewear configuration profiles, the one or more persona eyewear configuration profiles are based on the aesthetic component score, and an eyewear configuration is selected for the user based on the generated user eyewear configuration profile.

[0004] The foregoing paragraphs are provided as a general introduction and are not intended to limit the scope of the following claims.A fuller understanding of the described embodiments and further advantages will be obtained by reference to the following detailed description taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] A more complete appreciation of the present disclosure and its many attendant advantages will be readily obtained as the same become better understood by reference to the following detailed description when considered in conjunction with the accompanying drawings, in which:

[0006] Figure 1 An eyewear equipment recommendation system according to one or more aspects of the disclosed subject matter is presented;

[0007] Figure 2 An eyewear outfit recommendation workflow according to one or more aspects of the disclosed subject matter is presented;

[0008] Figure 3 presents a visual representation of an exemplary neural network in accordance with one or more aspects of the disclosed subject matter;

[0009] Figure 4 An exemplary workflow for determining the most popular eyewear features for one or more avatars in a pool of avatars according to one or more aspects of the disclosed subject matter is presented;

[0010] Figure 5 An exemplary workflow for identifying what avatar corresponds to a consumer in accordance with one or more aspects of the disclosed subject matter is presented;

[0011] Figure 6 An exemplary workflow capable of associating an eyewear equipment profile with a particular celebrity in accordance with one or more aspects of the disclosed subject matter is presented;

[0012] Figure 7 An exemplary workflow for automatically engaging consumers by notifying them of opinions of specific celebrities in accordance with one or more aspects of the disclosed subject matter is presented;

[0013] Figure 8 An algorithm flow chart illustrating a method for selecting eyewear equipment according to one or more aspects of the disclosed subject matter;

[0014] Fig. 9 is an algorithmic flow chart of a method for creating a vector representation of a training vocabulary according to one or more aspects of the disclosed subject matter;

[0015] Fig.10 is an algorithmic flow chart of a method for applying a neural network to one or more images according to one or more aspects of the disclosed subject matter;

[0016] Fig.11 is an algorithmic flow chart of a method for determining which statistics to output based on a use case in accordance with one or more aspects of the disclosed subject matter;

[0017] Fig.12 is an algorithm flow chart of a method for associating eyewear equipment with a character image according to one or more aspects of the disclosed subject matter;

[0018] Fig.13 An algorithm flow chart illustrating a method for associating a user with an eyewear equipment profile according to one or more aspects of the disclosed subject matter; and

[0019] Fig.14 is a hardware block diagram of a server according to one or more exemplary aspects of the disclosed subject matter. DETAILED DESCRIPTION

[0020] The following description in conjunction with the accompanying drawings is intended to be a description of the various embodiments of the disclosed subject matter, and is not necessarily intended to represent only (multiple) embodiments. In some cases, the description includes specific details for providing an understanding of the disclosed subject matter. However, it is apparent to those skilled in the art that the embodiments can be implemented without these specific details. In some instances, known structures and components can be shown in block diagram form to avoid blurring the conception of the disclosed subject matter.

[0021] References throughout the specification to "one embodiment" or "an embodiment" mean that a particular feature, structure, characteristic, operation, or function described in connection with the embodiment is included in at least one embodiment of the disclosed subject matter. Thus, the phrases "in one embodiment" or "in an embodiment" appearing in the specification do not necessarily refer to the same embodiment. Further, the particular features, structures, characteristics, operations, or functions may be combined in any suitable manner in one or more embodiments. Further, embodiments of the disclosed subject matter are intended to be able to and do cover modifications and variations of the described embodiments.

[0022] It must be noted that, as used in the specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly indicates otherwise. That is, as used herein, the words "a," "an," and the like have the meaning of "one or more," unless clearly indicated otherwise. Additionally, terms such as "first," "second," "third," and the like identify only one of the multiple parts, components, reference points, operations, and / or functions described herein, and as such do not necessarily limit the embodiments of the disclosed subject matter to any particular configuration or orientation.

[0023] Referring now to the drawings, in which like reference numerals designate identical or corresponding parts throughout the several views, Figure 1 An eyewear outfit recommendation system 100 (referred to herein as system 100 ) in accordance with one or more aspects of the disclosed subject matter is presented. System 100 may include a server 110 , a remote device 120 , and a database 140 communicatively coupled via a network 130 .

[0024] The server 110 may represent one or more servers communicatively coupled to the remote device 120 and the database 140 via the network 130. The server 110 may include a computer, a laptop, a smart phone, a tablet, a PDA, etc. The server 110 may include a processing circuit system configured to select an eyewear outfit for a user using a neural network to analyze fashion trends and recommend the eyewear outfit to the user based on matching the user to an avatar linked to a specific eyewear outfit, which in turn may be linked to the user, as further described herein.

[0025] Remote device 120 may represent one or more remote devices communicatively coupled to server 110 and database 140 via network 130. Remote device 120 may include a computer, laptop, smart phone, tablet, PDA, and the like.

[0026] Database 140 may represent one or more databases communicatively coupled to server 110 and remote device 120 via network 130 .

[0027] The network 130 may represent one or more networks that connect the server 110, the remote device 120, and the database 140. The network 130 may be a public network, such as the Internet, a private network, such as a local area network (LAN) or a wide area network (WAN) network, or any combination thereof, and may also include a public switched telephone network (PSTN) or an integrated services digital network (ISDN) subnetwork. The network 130 may also be wired, such as an Ethernet network or a USB port, or may be wireless, such as a cellular network, including wireless cellular systems such as EDGE, 3G, 4G, LTE / LTE-A, etc. The wireless network may also be Wi-Fi, Bluetooth, or any other known form of wireless communication.

[0028] It should be understood that throughout the specification, the terms "eyewear equipment" and "vision equipment" are interchangeable. In addition, the terms "photograph", "picture" and "image" are interchangeable. Further, the terms "user" and "consumer" are interchangeable.

[0029] In general, the system 100 may recommend eyewear equipment to a user based on eyewear equipment features associated with the user. For example, the user may be a client / customer and / or a potential client. In order to recommend eyewear equipment based on the eyewear equipment features of the user, the system 100 may detect one or more fashion trends and identify popular eyewear features of a predetermined persona based on the one or more fashion trends. The system 100 may then identify the eyewear equipment features of the user based on the predetermined persona corresponding to the user. Once the eyewear equipment features are identified, the system 100 may make eyewear equipment recommendations based on the eyewear equipment features of the user.

[0030] Making the best possible recommendations and the timing of making recommendations are critical to effectively recommending eyewear equipment. In this context, a technical solution to the technical problem may include automatically scoring all available frames in a retailer's catalog based on global trends or based on eyewear-specific trends extracted from general opinions, or concentrated on the opinions of a group of pre-selected influencers and / or celebrities, personalized for a specific user, or targeted at anyone who is sensitive to trends and fashion. The opinions of the group of pre-selected influencers and / or celebrities may correspond to any data source and / or media format including social media engagement (e.g., likes, comments, shares, etc.). In addition, the technical solution may include automatically attracting a consumer by notifying the consumer of the opinions of a particular celebrity. Further, the "opinion" of a celebrity may include the fact that in the photos in which the celebrity appears, the celebrity actually wears the same eyewear equipment that they recommend to the consumer.

[0031] More specifically, with respect to detecting fashion trends, the system 100 may receive a plurality of images, and the plurality of images may be, for example, images of one or more persons. The system 100 may calculate a score for each of the plurality of images based on an estimated trend level. In one aspect, the trend level may be based on the number of times a particular item and / or aesthetic component in the image appears in the plurality of images. For example, if the color red appears with a high frequency in all of the received plurality of images, the color red may have a high trend level.

[0032] In addition, the system 100 can determine whether the user requests eyewear equipment recommendations based on a synthesis of all detected trends. For this type of request, the system 100 can output global trend statistics in response to the request based on a synthesis of all trends. Further, the global trend statistics can be based on the scores calculated for each of the multiple images. Alternatively or in addition, the user can request eyewear equipment recommendations based specifically on eyewear trends. Therefore, the system 100 can output eyewear trend statistics, which can be based on the scores calculated for each of the multiple images. In addition, in one aspect, when the score is greater than a predetermined minimum score, the score can be merged into the global trend statistics as well as the eyewear trend statistics.

[0033] Further, when identifying popular eyewear features for a predetermined persona based on the one or more fashion trends, the system 100 may define or compute a set of personas based on a human-typical profile as further described herein, which may be based on, for example, a plurality of images received from publicly available media corresponding to a set of pre-selected celebrities or trendsetters. The system 100 may then associate each pre-selected celebrity or trendsetter with a matching persona in the persona database. For example, the association may be based on physical and situational features of the pre-selected celebrity or trendsetter. For each persona, the system 100 may associate an eyewear outfit with each persona in the persona set.

[0034] In addition, in response to the received one or more images being one or more images of the user, the system 100 can generate a profile for the user, associate the user with a matching persona, and associate the user with an eyewear outfit based on the matching persona. Thus, fully customized recommendations can be made for the user because the eyewear outfit recommendations will be based on trends that are attractive to the user and based on the user's fashion style.

[0035] Figure 2 An eyewear equipment recommendation workflow 200 according to one or more aspects of the disclosed subject matter is shown. The workflow 200 may include a fashion trend box 205, a user personalization box 210, an eyewear equipment trend profile box 215, a similarity analysis box 220 including a box 225 for attribute matching with a retailer catalog, and an eyewear equipment recommendation box 230.

[0036] Fashion trend frame 205 can include the product range of this fashion trend and the crowd range of this fashion trend.For example, can determine the trend aspect from the general angle including checking specific clothes and accessories (such as hats, jewelry, cosmetics, clothes, handbags, shoes, scarves, shapes, colors, etc.), or specifically from the angle of checking glasses equipment.Further, can determine these trends according to the opinion of a group of influential people or according to a common opinion (from such as feedback of the general population, they may have or do not have any reputation in the fashion industry).In other words, the product range can include a general range, which can correspond to the comprehensive user of requesting all fashion trends, and the crowd range can include a general group, which can correspond to common opinions, and a segmentation group, which can include, for example, a group of pre-selected trendsetters and / or celebrities' opinions.Alternatively or additionally, the segmentation group can include the people of a specific category, and the specific category can be associated with, for example, age, gender, social occupation category, etc.

[0037] In one embodiment, for example, the fashion trends in the fashion trend box 205 can be identified based on image processing. In this case, a neural network can be used for image processing.

[0038] Figure 3 A visual representation of an exemplary neural network 300 according to one or more aspects of the disclosed subject matter is shown. Neural network 300 can receive a photo 305 as an input. For example, photo 305 can be a photo of one or more people. In addition, neural network 300 can include convolutional layer 310 and fully connected layer 315. Fully connected layer 315 can include one or more layers, for example, including L layer 320 to N-1 layer 325. In addition, fully connected layer 315 can include output layer 330.

[0039] First, neural network 300 may receive photos to be tagged. The photos to be tagged may be photos including one or more people (such as photo 305), and these photos may be photos of the faces and / or entire bodies of the people. In addition, the people in the photos may or may not wear eyeglasses. Further, one or more of the multiple photos may specifically be photos of eyeglasses.

[0040] For a tagging process specifically for detecting eyewear trends, eyewear in photos of people wearing eyewear can be tagged. For example, a frame manufacturer, lens manufacturer, etc. can label the eyewear to describe its shape, thickness, frame material, color, lens color, mirror effect, or brand of the eyewear.

[0041] For the labeling process for detecting general trends, objective information and subjective information can be collected to label each image. For example, for each image, anyone can be asked to provide objective information to label the image, and people from the fashion industry can be asked to label the image for subjective information, for example, based on their expertise. Objective information can include information about the person in the picture, including gender, approximate age, hair color, eye color, skin color, etc. In addition, the objective information can include situational tags about the person, including what the person is doing, the location of the person in the photo (such as outdoor, indoor, beach, commuting, etc.). Subjective information can include the style of the person in the image with tags including charm, fashionable, serious, etc. In this case, it should be preferred to be labeled by professionals from the fashion industry. In addition, the subjective information can include tags of clothing and accessories worn by the person, including the type, color, shape, brand, pattern, texture, product category, details, and features of items (including hats, jewelry, cosmetics, clothes, handbags, shoes, etc.).

[0042] After labeling is completed, the labeled images can be used to build a deep learning database. The deep learning database can correspond to one of the one or more databases 140, for example. Next, the deep learning database can be used to train the deep learning algorithm by using the convolution layer 310 to extract feature maps from each labeled image, and using the fully connected layer 315 to extract vectors of attributes from the labeled images. Once trained successfully, the algorithm can be used to take any picture of a person as input, the picture only including the person's face or showing their entire body, and the neural network 300 can provide various information, including information about the person in the picture (e.g., gender, approximate age, hair color, eye color, skin color, etc.), the person's scene including activities and locations (e.g., outdoors, indoors, beach and / or commuting), the person's style (e.g., charm, fashionable, serious, etc.), the clothing and accessories worn by the person (including one or more of the type, color, shape, brand, pattern, texture, product category, details and features of hats, jewelry, cosmetics, clothes, handbags and shoes), and the attributes of the visual equipment worn by the person. In addition, the information provided by the neural network 300 can include confidence.

[0043] In one embodiment, another or several other neural networks can be used to perform text mining (e.g., word2vec) on text information from social media or the user. In order to build a word and / or tag training corpus, a vocabulary can be defined to include eyewear equipment trend detection and general trend detection as described herein with respect to labeled images. After defining the vocabulary, the one or more neural networks can be used to create a vector representation of the training vocabulary. Using the vector representation of the training corpus, any text can be received and an algorithm (e.g., a k-nearest neighbor algorithm) can be used to detect information including information about the person in the picture (e.g., gender, age approximation, hair color, eye color, skin color, etc.), the person's scene including activities and locations (e.g., outdoors, indoors, beach and / or commuting), the person's style (e.g., glamour, fashionable, serious, etc.), the clothing and accessories worn by the person (including one or more of the type, color, shape, brand, pattern, texture, product category, details and features of hats, jewelry, cosmetics, clothes, handbags and shoes), and the attributes of the visual equipment worn by the person. In addition, the information provided by word2vec can include confidence.

[0044] Back to Figure 2, the trend detection block 205 may receive input. Regarding the source of the input, a plurality of images may be provided to the neural network 300, as has been described herein. More specifically, the plurality of images may be an open source image / video pool of any person, or the open source image / video pool may be a set of pre-selected celebrities and / or trendsetters (e.g., with a certain amount of followers), which may come from one or more of public social media profiles, community reviews, fashion magazines, etc.

[0045] In addition, fashion trend frame 305 can be configured for the selection and scoring of fashion trends. For example, these inputs are selected within a specific time frame (for example, 6 months) consistent with the frequency of trend changes in the fashion industry. Score (such as aesthetic component score) can be calculated for it according to the estimated trend level of each selected input. The aesthetic component score can be based on the number of occurrences of the project (for example, type, color, shape, product, pattern, brand, product category, etc.) in a text, a picture or a video, or several texts, several pictures or several videos. The aesthetic component score can be weighted according to some criteria, which include the time when the picture / text is published, the time period when all inputs related to the project are published, the person who publishes the input (for example, trend influencer / celebrity can have a higher weight), feedback from people about the input (for example, like number, like number, positive comment, forwarding number, etc.), and the previous score of the same aesthetic component. As a result, the trend level can be given according to the score of each input. In addition, in one aspect, only the input with good evaluation will be processed. For example, the aesthetic component score must be greater than a predetermined minimum score (for example, score>7 / 10).

[0046] In addition, the manner in which this information is communicated through the neural network 300 depends on the use case. For example, for a first use case, the user may want eyewear outfit recommendations based on global trends. In this case, global statistics can be calculated based on various features detected by the neural network 300 on the selected image, or detected by semantic analysis on the selected text. The various features used to calculate global statistics can include: clothes and accessories worn by the person (e.g., hats, jewelry, makeup, clothes, handbags, shoes, etc.), including type, color, shape, brand, pattern, texture, product category, details and features.

[0047] In another use case, a user may want recommendations for eyewear gear specifically based on eyewear trends. In this case, eyewear statistics may be calculated based on various features detected by the neural network 300 on the selected images, or detected through semantic analysis on the selected text. For example, the various features may include attributes of the eyewear gear worn by the people in the images.

[0048] The output of the fashion trends box 305 may output statistics describing the detected trends (e.g., based on the use case). Additionally, these statistics may take into account trend levels calculated based on aesthetic component scores as described herein. With respect to the first use case where the user wants eyewear outfit recommendations based on global trends, the global statistics output may include popular colors (e.g., red is detected as popular with a trend level of 9 / 10), popular shapes (e.g., round is detected as popular with a trend level of 8 / 10), popular materials (e.g., leather is detected as popular with a trend level of 8 / 10), etc. Regarding the use case where a user wants recommendations for eyewear outfits specifically based on eyewear trends (e.g., frames, lenses, etc.), these eyewear trend statistics may include popular eyewear colors (e.g., red frames detected as popular with a trend level of 9 / 10), popular eyewear thicknesses (e.g., nylor glasses detected as popular with a trend level of 7 / 10), popular eyewear shapes (e.g., round frames detected as popular with a trend level of 8 / 10), popular eyewear materials (e.g., cellulose acetate frames detected as popular with a trend level of 8 / 10), popular eyewear textures (e.g., tortoise shell frames detected as popular with a trend level of 7.5 / 10), popular eyewear effects (e.g., lenses with purple reflections detected as popular with a trend level of 8.5 / 10), etc.

[0049] With regard to the maintenance of the trend detection box 305, the trend detection box 305 can be updated regularly (e.g., once every six months) to capture the rapid trend evolution in the fashion industry. For example, every six months, all previously calculated statistics can be cleared, the input data (e.g., images or text) can be updated by clearing the previously selected input data, and the data source (e.g., public profile social media, community comments, fashion magazines, etc.) can be reviewed again to select new input data, and new statistics (e.g., global statistics and glasses trend statistics) can be calculated from the newly selected input data. In one aspect, trend information from previous cycles can be retained to compare with trend information of new cycles, and can be used to incorporate it into statistics of new cycles. For example, in the case where there is a clear overlap in the trend (e.g., red is still a popular color), the weight of red can be greater because red is a strong long-term trend.

[0050] The user personalization frame 210 can handle the personalized aspects of the eyewear equipment recommendation. For example, the server 110 can be configured to define or calculate a set of typical personas (e.g., gender, age category, style, favorite brands, lifestyle, etc.). A typical person can correspond to a typical human profile (e.g., "a twenty-something urban girl who likes high-end brands"). After defining or calculating this set of typical personas, the server 110 can be configured to associate the eyewear equipment with the user personalization frame 210. Figure 4Each character in the game is associated with the character.

[0051] Figure 4 An exemplary workflow 400 for determining the most popular eyewear features of one or more personas (e.g., persona 405 and persona 410) in a persona pool 415 according to one or more aspects of the disclosed subject matter is shown. To associate eyewear equipment with each persona, a neural network 300 can be applied to each image / video in a pool of open source images / videos of the set of pre-selected celebrities or trendsetters (e.g., input 420). By running the neural network 300 on each image / video in the pool (the images / videos generally include face, style, situation, and eyewear equipment analysis), various information about one or more people in the images / videos can be obtained. For example, the output of the neural network 300 may provide information about the person in the picture (e.g., gender, approximate age, hair color, eye color, skin color, etc.), the context of the person, including what they are doing and where they are (e.g., outdoors, indoors, at the beach, commuting, etc.), the person's style (e.g., glamour, hipster, serious, etc.), the clothing and accessories worn by the person (e.g., hats, jewelry, makeup, clothes, handbags, shoes, etc.), including type, color, shape, brand, pattern, texture, product category, details and features, and the attributes of the eyewear equipment worn by the person (including shape, material, color, texture, thickness, iconic brand, etc.). In other words, the output of the neural network 300 may be used to determine the best matching persona (e.g., as represented by arrow 425), and the corresponding eyewear profile of the person that matches the persona may be fed to the visual equipment profiles 435a, 435b corresponding to the persona (e.g., as represented by arrow 430).

[0052] Next, each person in the picture can be associated with the character image that is most similar to him / her (e.g., arrow 425). For example, the person will be associated with the character image that has the most similar features to him / her, and the features can be based on the information about the person in the picture output by the neural network 300. In addition, when associating the person with the character image, different weights W1, W2, ..., W N Different features may be assigned (eg, hair color may have a weight of 0.05 compared to style, whereas style may have a weight of 0.2 because style is detected as more discriminative than hair color from a fashion perspective).

[0053] Then, for each persona, server 110 may be configured to determine the most popular eyewear features for that persona to generate a pool of personas that are linked to eyewear outfit profiles 440. For example, the attributes of the eyewear worn by each person in the pool of personas may be linked to the associated persona, and as further described herein, the attributes may be weighted according to their calculated level of matching with the persona.

[0054] Referring again to the user personalization block 210, the server 110 may be configured to identify which avatar corresponds to the consumer who desires an eyewear outfit recommendation based on the pool of avatars linked to the visual outfit profile 440, such as Figure 5 as shown in .

[0055] Figure 5 An exemplary workflow 500 for identifying what persona corresponds to a consumer according to one or more aspects of the disclosed subject matter is shown. For example, a consumer may upload pictures and / or videos of themselves that represent them in situations they like. Alternatively or additionally, the input may be a questionnaire filled out by the consumer, such as a questionnaire including lifestyle questions. If the consumer uploads pictures and / or videos, the same process described above for matching an open source image / video pool about a set of pre-selected celebrities or trendsetters may be applied to the pictures and / or videos of the consumer provided by the consumer. Similarly, if the consumer fills out a questionnaire, the consumer may be associated with a persona, such as based on a word2vec process. Thus, the workflow 500 may determine the most popular eyewear attributes for the consumer by identifying the persona that most closely matches the consumer and linking the eyewear attributes associated with the persona to the consumer.

[0056] In other words, reference Figure 5, a consumer may provide a photo 505 as an input to a neural network 300, which may perform facial analysis, situational analysis, style analysis, etc. as described herein. The neural network 300 may output a consumer persona profile 510 that includes all of the information identified in the consumer photo 505. The consumer persona profile 510 may be used to match the features included in the consumer persona profile with one or more personas in the persona pool linked to the eyewear equipment profile 440. The matching process may result in a best match (such as match 515). For example, match 515 may correspond to the consumer having the best match with "persona 2". Then, as a result of the consumer being most closely matched to "persona 2", the eyewear equipment profile associated with "persona 2" may be linked to the consumer, which may correspond to the eyewear equipment trend profile 215. In other words, the eyewear equipment trend profile 215 of the consumer may correspond to the most popular eyewear equipment for the consumer.

[0057] Reference again Figure 2 , the similarity analysis block 220 may be configured to match the attributes of the eyewear equipment trend profile 215 with frames in the retailer catalog (i.e., attribute matching block 225). In other words, the similarity analysis block 220 may be configured to tailor the eyewear equipment recommendation to the eyewear equipment available in one or more specific retailer catalogs. For example, based on the one or more personalized attributes identified from the trend detection block 205 or the user personalization block 210 and the trend level of the personalized attributes of the eyewear equipment, the server 110 may be configured to find a list of eyewear equipment corresponding to these attributes from the retailer catalog.

[0058] For example, a process for finding and displaying a list of eyewear gear from a retailer catalog may include determining attributes of eyewear available in the retailer catalog, associating the most popular eyewear attributes for the consumer (e.g., eyewear gear trend profile 215) with specific eyewear available in the retailer catalog, and displaying a selection of eyewear gear to the user.

[0059] Determining the attributes of the eyewear equipment available in the retailer catalog may include identifying the shape, material, color, texture, thickness, iconic brand, etc. of the eyewear equipment. The attributes of the eyewear equipment in the retailer catalog may be provided manually (e.g., by a person), looked up in the retailer catalog, or scraped from the Internet. However, these attributes are preferably outputs of a neural network 300 that takes as input a pool of images of eyewear equipment in the retailer catalog.

[0060] Associating the most popular eyewear attributes for the consumer with the specific eyewear available in the retailer's catalog may include maximizing their common features (e.g., shape, material, color, texture, thickness, iconic brand, etc.). Additionally, maximizing their common features may include considering different weights W1, W2, ... W N , these features may be given different weights depending on the situation. For example, style may have a weight of 0.6, material may have a weight of 0.2, and so on. Style may have a larger weight because style may be detected as a highly discriminative parameter in the input. Alternatively or additionally, the retailer may have decided that style should be more important than other features. Each trend may be understood as a specialization of a given feature. For example, red is a specialization of the feature "color". Therefore, each trend may have a trend level T 1,1 ,...,T i,j ,...T N,M For example, the red frame (i.e., the specialization j(red) of feature i(color)) is detected as having a trend level T i,j =9 / 10). The weight and trend level are then combined to calculate the expected level for each attribute:

[0061] D i,j =W 特征i *T 趋势i,j

[0062] Equation 1

[0063] For example, since the color feature has been given a weight of 0.4, and red is popular and has been given a trend level of 0.9, then, applying Equation 1, the expected level for red is: 红色 =0.4*0.9=0.36

[0064] Therefore, a global trend level is calculated for each eyewear equipment based on its attributes and the calculated desired level. Equation 1 can be an average of the desired levels for each attribute of the eyewear equipment. The eyewear equipment with the best global trend level can be considered popular for the consumer and selected, and then additional selections can be made based on other criteria including promotions, inventory (i.e., availability), price, number of displayed items, etc., depending on the needs of the retailer. Finally, the selected glasses can be displayed to the user / consumer, for example, via the remote device 120. Displaying the selected glasses can correspond to the eyewear equipment recommendation frame 230.

[0065] In one embodiment, the server 110 may be configured to automatically attract a consumer by notifying the consumer of a particular celebrity's opinion. For example, attracting a consumer based on a particular celebrity's opinion may be performed in two ways. In the first way, the server 110 may be configured to remind the consumer that the consumer's morphology is similar to that of a particular celebrity, and to suggest to the consumer an eyewear outfit similar to that currently worn by the particular celebrity. In the second way, the server 110 may be configured to remind the consumer that a particular celebrity wears eyewear outfit similar to that currently being viewed by the consumer.

[0066] Figure 6An exemplary workflow 600 capable of associating an eyewear equipment profile with a specific celebrity according to one or more aspects of the disclosed subject matter is shown. In order to automatically attract consumers by notifying consumers of the opinions of specific celebrities, the server 110 can be configured to define a set 605 of specific celebrities who are considered trendsetters. Determining whether a celebrity is a trendsetter can be done using objective criteria, for example, including the number of followers and / or friends on a social media platform. Next, the server 110 can be configured to build a database (e.g., database 140) of open source images / videos about the set 605 of the specific celebrity, which can come from one or more of a public social media profile, community comments, fashion magazines, etc. Using the set 605 of the specific celebrity as input, eyewear equipment can be associated with each celebrity by applying the neural network 300 to the set 605 of the specific celebrity. In general, the neural network 300 can perform facial analysis, style analysis, situational analysis, eyewear equipment analysis, and the like. For example, the server 110 may be configured to run the neural network 300 on each image / video constituting the set 605 of the particular celebrity to determine, based on the image / video, information about the celebrity (e.g., gender, approximate age, hair color, eye color, skin color, etc.), the context of the celebrity, including what the celebrity is doing and where the celebrity is (e.g., outdoors, indoors, at the beach, commuting, etc.), the style of the celebrity (e.g., glamour, trendy, serious, etc.), clothing and accessories worn by the celebrity (e.g., hats, jewelry, makeup, clothes, handbags, shoes), including type, color, shape, brand, pattern, texture, product category, details and features, and attributes of the eyewear equipment worn by the celebrity, including shape, material, color, texture, thickness, iconic brand, etc. Each celebrity may then be associated with one or more eyewear equipment attribute lists. For example, because the identity of the celebrity may be known, each celebrity may be linked to one or more eyewear equipment attribute lists. For example, the output of the neural network 300 corresponding to the face, style, and situational analysis may be fed to information corresponding to a particular celebrity (e.g., as indicated by arrow 615), and the output of the neural network 300 corresponding to the eyewear equipment analysis may be fed to an eyewear equipment profile corresponding to the celebrity (e.g., as indicated by arrow 610). Thus, the output of the neural network 300 may be a celebrity database linked to a visual equipment profile 620. The celebrity database linked to the visual equipment profile 620 may, for example, be one of the one or more databases 140. More specifically, the celebrity database linked to the visual equipment profile 620 may include morphological information and situational information about the celebrity and associated eyewear equipment attributes. The manner in which the database 620 is used depends on the use case, such as Figure 7 shown.

[0067] Figure 7 An exemplary workflow 700 for automatically attracting a consumer by notifying the consumer of a particular celebrity's opinion in accordance with one or more aspects of the disclosed subject matter is shown. For example, in a use case where the consumer is reminded that the consumer's morphology is similar to that of a particular celebrity, the server 110 may be configured to recommend one or more eyewear devices that are similar to the eyewear device currently worn by the particular celebrity. The server 110 may receive one or more photos and / or videos (e.g., consumer photos 705) of the consumer wearing or not wearing the eyewear device. The one or more photos and / or videos may include photos and / or videos of the consumer's face and / or the consumer's entire body. A neural network (e.g., neural network 300) may be used to compute information from the consumer photo 705, including information about the consumer based on the image / video (e.g., gender, approximate age, hair color, eye color, skin color, etc.), the context of the consumer, including what the consumer is doing and where the consumer is (e.g., outdoors, indoors, at the beach, commuting, etc.), the style of the consumer (e.g., glamour, hipster, serious, etc.), clothing and accessories worn by the consumer (e.g., hats, jewelry, makeup, clothes, handbags, shoes), including type, color, shape, brand, pattern, texture, product category, details and features, and attributes of the eyewear equipment worn by the consumer, including shape, material, color, texture, thickness, iconic brand, etc. In addition, the server 110 may be configured to browse the database 620 and compute one or more matching coefficients (e.g., matching coefficients 710a, 710b) of the consumer with each celebrity relative to the identified information about the consumer based on using the consumer photo 705 as input and based on output from the neural network 300.

[0068] For example, a matching coefficient may be calculated for each standard category, the criteria including general information (e.g., gender, age, race, etc.), morphological information (e.g., face shape, hair color, etc.), situational information (e.g., location, activity), style information, clothing, accessory information (e.g., cosmetic accessories), etc. For example, matching coefficient 710a has a very high morphological coefficient (e.g., a morphological coefficient of 0.92) with celebrity 715, which may indicate a strong match. For each category, server 110 may be configured to find a celebrity with the highest matching coefficient with the consumer, and retrieve (multiple) visual equipment attributes linked to each corresponding celebrity. Thus, match 720 may correspond to the consumer having the best match with celebrity 715, and the eyewear equipment profile associated with celebrity 715 may be linked to the consumer. Subsequently, server 110 may be configured to output recommendation 725, which corresponds to recommending one or more eyewear equipment features for the consumer based on match 720.

[0069] In addition, recommendation 725 may include additional reminders to further engage the consumer. A first example includes being output as a recommendation or with a recommendation: "Wow! Have you noticed that you look like Ryan Gosling. He's wearing these frames now. What do you think?" The reminder may further include one or more of a photo of Ryan Gosling and a photo of the frames mentioned in the recommendation. As a second example, the recommendation may include "Hi! You seem to like Gothic looks, like Madonna. She was recently seen wearing these sunglasses. What do you think?" The reminder may further include one or more of a photo of Madonna and a photo of the sunglasses mentioned in the recommendation. As a third example, the recommendation may include "Hi! You seem to appreciate big earrings, like Beyoncé. She's also wearing sunglasses with them. What do you think?" The reminder may further include one or more of a photo of Beyoncé and a photo of the sunglasses mentioned in the recommendation.

[0070] Regarding the second use case corresponding to reminding the consumer that a particular celebrity wears eyewear equipment similar to the eyewear equipment currently being browsed by the consumer, for example, the input may include attributes of a particular eyewear equipment (e.g., the eyewear equipment currently being browsed by the consumer). The server 110 may be configured to browse the database 620 and calculate a matching coefficient between the eyewear equipment being browsed by the consumer and the eyewear equipment in the database 620. In addition, the server 110 may be configured to find the eyewear equipment with the highest matching coefficient with the eyewear equipment being browsed by the consumer and retrieve the celebrities associated with the eyewear equipment from the database 620. Based on finding the eyewear equipment that best matches the eyewear equipment of these celebrities in the database 620, the server 110 may be configured to remind the consumer who has retrieved the celebrity who recently wore the eyewear equipment similar to the eyewear equipment currently being browsed by the consumer. For example, the reminder may include "Did you know that Eva Longoria is currently wearing frames that are very similar to the frames you are currently trying on?"

[0071] Figure 8 An algorithm flow chart of a method for selecting eyewear equipment according to one or more aspects of the disclosed subject matter is presented.

[0072] In S805, the server 110 may receive one or more images. The one or more images may include one or more people. For example, as further described herein, the one or more images may be an open source image / video pool of any person or a group of pre-selected celebrities or trendsetters. In addition, the received one or more images may be images of consumers.

[0073] In S810, a neural network (eg, neural network 300) may be applied to the one or more images. Neural network 300 may classify at least one aesthetic component of each of the one or more images.

[0074] In S815, the server 110 may generate an aesthetic component score for each of the one or more images. For example, the aesthetic component score may be based on the number of occurrences of the aesthetic component. In addition, the aesthetic component score may be calculated to take into account the weight of the aesthetic component (for example, if an aesthetic component is considered to be more relevant to fashion, it may have a higher weight).

[0075] In S820, the server 110 may generate a user eyewear equipment profile for the user. For example, the user may be matched with a character image from a character image database (eg, database 440).

[0076] In S825, the server 110 may match the user with a persona from the persona database. In other words, each persona in the persona database may be linked to one or more persona eyewear outfit profiles. The user may then be linked to the eyewear outfit configuration file of the persona that most closely matches the user. Additionally, the one or more persona eyewear outfit profiles may be based on the aesthetic component score.

[0077] In S830, the server 110 may be configured to select eyewear equipment for the user based on the user eyewear equipment profile. After selecting eyewear equipment for the user, the process may end.

[0078] Fig. 9 is an algorithmic flow chart of a method for creating a vector representation of a training vocabulary in accordance with one or more aspects of the disclosed subject matter.

[0079] In S905 , the server 110 may receive text corresponding to the usage description.

[0080] In S910 , the server 110 may train a vocabulary corresponding to the received text.

[0081] In S915, the server 110 may create a vector representation of the training vocabulary, for example, using word2vec. After creating the vector representation of the training vocabulary, the process may end.

[0082] Fig.10 is an algorithm flow chart of a method for applying a neural network to one or more images according to one or more aspects of the disclosed subject matter. Fig.10 In the embodiment, steps 1005 to 1025 may correspond to S810.

[0083] In S1005 , the server 110 may be configured to sum the number of occurrences of each classified aesthetic component in one or more images.

[0084] In S1010, the server 110 may apply a weight to each of the classified aesthetic components. For example, the number of occurrences of each of the classified aesthetic components may be weighed according to predetermined criteria, including the time when the image in which the item appears was published, the time period during which all the multiple images of the item were published, the person who provided the image, and feedback from others about the image (including social media participation).

[0085] In S1015 , the server 110 may calculate an aesthetic component score for each of the one or more images based on the frequency at which each aesthetic component appears in the one or more images.

[0086] In S1020, the server 110 may determine whether the end of a predetermined fashion interval has been reached. For example, based on fashion trends, the predetermined fashion interval may be six months. If it is determined that the end of the fashion trend interval has not been reached, the process may continue to check in S1020 whether the end of the predetermined fashion interval has been reached. However, in response to determining that the end of the fashion trend interval has been reached, the server 110 may receive another set of one or more images in S1025.

[0087] In S1025, when the end of the fashion interval has been reached, the server 110 may receive another set of one or more images. After receiving the another set of one or more images, the process may end.

[0088] Fig.11 is an algorithmic flow chart of a method for determining which statistics to output based on a use case in accordance with one or more aspects of the disclosed subject matter.

[0089] In S1105, the server 110 may be configured to determine whether the user has requested a comprehensive eyewear equipment recommendation based on all aesthetic components. In response to determining that the user has requested a comprehensive eyewear equipment recommendation based on all aesthetic components, global statistics may be output in S1110. The global statistics may be based on the trend level of all aesthetic components, for example, as described herein. After outputting the global statistics, the process may end.

[0090] However, in response to determining that the user has not requested comprehensive eyewear equipment recommendations based on all aesthetic components, the server 110 may be configured to determine whether the user has requested eyewear equipment recommendations based on eyewear aesthetic components in S1115 .

[0091] In response to determining that the user has not requested eyewear recommendations based on eyewear aesthetic components in S1115, the process may end. However, in response to determining that the user has requested eyewear equipment recommendations based on eyewear aesthetic components, the server 110 may output eyewear statistics in S1120. The eyewear statistics may be based on, for example, aesthetic features of the eyewear identified in one or more received images. After outputting the eyewear statistics, the process may end.

[0092] Fig.12 is an algorithm flow chart of a method for associating eyewear equipment with a character image according to one or more aspects of the disclosed subject matter.

[0093] In S1205, the server 110 may be configured to extract attributes from the received one or more images. For example, the one or more images may be from publicly available media corresponding to a set of pre-selected celebrities or trendsetters.

[0094] In S1210, the server 110 may generate a character image database by defining or calculating a character image set. For example, each character image in the character image database may have a human-typical profile. The human-typical profile may be based on physical features and situational features of the person in the one or more images, which may be identified by applying the neural network 300 to the one or more images.

[0095] In S1215, the server 110 may associate each pre-selected celebrity or trendsetter with a matching character image in the character image database. The association may be based on the physical features and situational features of the pre-selected celebrity or trendsetter. In other words, the pre-selected celebrity or trendsetter is associated with a human typical character image that most closely matches the appearance of the celebrity or trendsetter.

[0096] In S1220, the server 110 may associate eyewear equipment with each persona in the persona set based on the aesthetic component scores of the eyewear equipment of the pre-selected celebrity or trendsetter. For example, the associated eyewear equipment may correspond to one or more persona eyewear equipment profiles. In other words, each persona in the persona set may be linked to a specific eyewear equipment and / or eyewear equipment feature. After associating eyewear equipment with each persona in the persona set, the process may end.

[0097] Fig.13 An algorithm flow chart of a method for associating a user with an eyewear equipment profile according to one or more aspects of the disclosed subject matter is presented.

[0098] In S1305, the server 110 may generate a profile for the user based on the physical features and situational features of the user. For example, the server may generate a profile in response to the received one or more images being one or more images of the user.

[0099] In S1310 , server 110 may associate the user with a matching avatar in the avatar database that most closely matches the profile generated for the user based on the user's physical features and situational features.

[0100] In S1315, the server 110 may associate the user with a user eyewear equipment profile based on the matched character image. For example, the user eyewear equipment profile may be based on one or more character image eyewear equipment profiles of the matched character image. In other words, the user may be associated with the character image that most closely matches the user, and then the eyewear equipment linked to the character image may be linked to the user. After the user is associated with the user eyewear equipment profile based on the matched character image, the process may end.

[0101] On the top Figures 8 to 13 In the description, any process, description or box in the flowchart can be understood as a module, segment or portion representing a code, which includes one or more executable instructions for implementing a specific logical function or step in the process, and alternative implementations are included within the scope of the exemplary embodiments of the present invention, wherein the functions can be performed in an order different from the order shown or discussed, including substantially simultaneously or in reverse order, depending on the functions involved, as will be understood by those skilled in the art. The various elements, features and processes described herein can be used independently of each other, or can be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of this disclosure.

[0102] The system 100 includes several advantages, including providing a more accurate recommendation of eyewear equipment to a consumer. Additionally, the recommendation may include an increase in consumer confidence that the eyewear equipment will be popular and also a good match to the consumer's style. Thus, the recommendation may be more specifically targeted to the user. Further, in addition to eyewear-specific trends, the eyewear equipment may also be matched to the consumer based on what celebrities and trendsetters are currently wearing and general trends. Thus, the technical problem of effectively selecting eyewear equipment for a consumer may be solved by a technical solution of automatically selecting eyewear equipment for a user using a neural network to analyze fashion trends and recommend the eyewear equipment to the user based on matching the user with a persona that is linked to a specific eyewear equipment, which persona may then be linked to the user.

[0103] Next, refer to Fig.141 to describe the hardware description of the server 110 according to an exemplary embodiment. The hardware description described herein may also be a hardware description of the remote device 120 and / or the processing circuit system. Fig.14 , the server 110 includes a CPU 1400 that performs one or more of the above / below processes. Process data and instructions can be stored in a memory 1402. These processes and instructions can also be stored on a storage medium disk 1404 such as a hard disk drive (HDD) or a portable storage medium, or can be stored remotely. Further, the claimed improvements are not limited by the form of the computer-readable medium storing the instructions of the processes of the present invention. For example, the instructions can be stored on a CD, DVD, flash memory, RAM, ROM, PROM, EPROM, EEPROM, hard disk, or any other information processing device (such as a server or computer) with which the server 110 communicates. In addition, for example, a graphics processing unit (GPU) 1428 (such as an NVIDIA GeFroce RTX 2080Ti) can handle some or all of the neural network processing.

[0104] Further, the claimed advances may be provided as a utility application, a background daemon, or a component of an operating system, or a combination thereof, executed in conjunction with a CPU 1400 and an operating system such as Microsoft Windows, UNIX, Solaris, LINUX, Apple MAC-OS, and others known to those skilled in the art.

[0105] The hardware elements used to implement the server 110 may be implemented by various circuit system elements. Further, each of the functions of the above-mentioned embodiments may be implemented by a circuit system including one or more processing circuits. Fig.14 As shown, the processing circuitry includes a specially programmed processor, such as processor (CPU) 1400. The processing circuitry also includes devices such as application specific integrated circuits (ASICs) and conventional circuit components arranged to perform the functions described.

[0106] exist Fig.14 In the embodiment, the server 110 includes a CPU 1400 that performs the above process. The server 110 can be a general purpose computer or a specific dedicated machine. In one embodiment, when the processor 1400 is programmed to perform eyewear equipment selection (and in particular, reference Figures 8 to 13 When performing any of the processes discussed herein, server 110 becomes a specific, dedicated machine.

[0107] Alternatively or additionally, as will be appreciated by those skilled in the art, CPU 1400 may be implemented on an FPGA, ASIC, PLD or using discrete logic circuits. Further, CPU 1400 may be implemented as multiple processors working in parallel to execute instructions of the above-described inventive process.

[0108] Fig.14 The server 110 in the embodiment also includes a network controller 1406, such as an Intel Ethernet PRO network interface card from Intel Corporation of the United States, for interfacing with the network 130. As can be appreciated, the network 130 can be a public network, such as the Internet, or a private network, such as a LAN or WAN network, or any combination thereof, and can also include a PSTN or ISDN subnetwork. The network 130 can also be wired, such as an Ethernet network, or can be wireless, such as a cellular network, including EDGE, 3G and 4G wireless cellular systems. The wireless network can also be WiFi, Bluetooth or any other known wireless communication form.

[0109] The server 110 further includes a display controller 1408, such as a graphics card or graphics adapter, for connecting to a display 1410 (such as a monitor). The general I / O interface 1412 is connected to a keyboard and / or mouse 1414, and a touch screen panel 1416 on or separate from the display 1410. The general I / O interface is also connected to various peripheral devices 1418, including printers and scanners.

[0110] The sound controller 1420 may also be provided in the server 110 to be connected to the speaker / microphone 1422 to provide sound / music.

[0111] The general storage controller 1424 connects the storage medium disk 1404 to the communication bus 1426, which may be ISA, EISA, VESA, PCI or the like, to interconnect all components of the server 110. For the sake of brevity, descriptions of the general features and functions of the display 1410, keyboard and / or mouse 1414, and display controller 1408, storage controller 1424, network controller 1406, sound controller 1420, and general I / O interface 1412 are omitted here because these features are known.

[0112] The exemplary circuit elements described in the context of this disclosure may be replaced by other elements and may be constructed differently from the examples provided herein. In addition, circuits configured to perform the features described herein may be implemented in multiple circuit units (e.g., chips), or these features may be combined in circuits on a single chipset.

[0113] The functions and features described herein may also be performed by various distributed components of the system. For example, one or more processors may perform these system functions, wherein the processors are distributed on a plurality of components that communicate with a network. In addition to various human-computer interaction devices and communication devices (e.g., display monitors, smart phones, tablet computers, personal digital assistants (PDAs)), distributed components may also include one or more clients and server machines that may share processing. The network may be a dedicated network, such as a LAN or WAN, or may be a public network, such as the Internet. Input to the system may be received via direct user input, and may be received remotely in real time or as a batch process. Additionally, some implementations may be performed on modules or hardware different from the described modules or hardware. Therefore, other implementations are within the scope that may be claimed.

[0114] Having now described embodiments of the disclosed subject matter, it will be apparent to those skilled in the art that the foregoing is merely illustrative and non-restrictive and is presented only as an example. Therefore, although specific configurations have been discussed herein, other configurations may also be adopted. The present disclosure implements many modifications and other embodiments (e.g., combinations, rearrangements, etc.), and these modifications and other embodiments are within the scope of those of ordinary skill in the art and are intended to belong to the scope of the disclosed subject matter and any equivalents thereof. The features of the disclosed embodiments may be combined, rearranged, omitted, etc. within the scope of the present invention to produce additional embodiments. In addition, certain features may sometimes be used to obtain advantages without the need to use other features accordingly. Therefore, the applicant intends to include all these substitutions, modifications, equivalents, and variations within the spirit and scope of the disclosed subject matter.

[0115] Embodiments of the present disclosure may also be described as follows in brackets.

[0116] (1) A server comprising: a processing circuit system configured to: receive one or more images, the one or more images including one or more representations of a person, apply a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, generates an aesthetic component score for each of the one or more images, generates a user eyewear configuration profile for a user, the user is matched with a persona from a persona database, each persona in the persona database is linked to one or more persona eyewear configuration profiles, the one or more persona eyewear configuration profiles are based on the aesthetic component score, and select eyewear configuration for the user based on the generated user eyewear configuration profile.

[0117] (2) A server according to (1), wherein the processing circuit system is further configured to: sum the number of occurrences of each classified aesthetic component in the one or more images, wherein the number of occurrences is weighed according to predetermined criteria, the predetermined criteria including one or more of the following: the time when the image in which the item appears was published, the time period in which all of the multiple images of the item were published, the person who provided the image, and feedback from others about the image, including social media engagement, and calculate an aesthetic component score for each of the one or more images based on the frequency with which each aesthetic component appears in the one or more images, wherein each aesthetic component is weighted based on the one or more predetermined criteria and each aesthetic component in each of the one or more images is combined to calculate aesthetic component statistics for the one or more images.

[0118] (3) A server according to (1) or (2), wherein the processing circuit system is further configured to: determine whether the user's eyewear equipment selection is based on a combination of all aesthetic components in the one or more images, output global statistics in response to determining that the user's eyewear equipment selection is based on the attributes of all aesthetic components in the one or more images, the global statistics based on the aesthetic component score, determine whether the user's eyewear equipment selection is based on the eyewear aesthetic components in the one or more images, and output eyewear statistics in response to determining that the user's eyewear equipment selection is based on all aesthetic components in the one or more images, the eyewear statistics based on the aesthetic component score, wherein the aesthetic component score is merged into the global statistics and the eyewear statistics only when the aesthetic component score is greater than a predetermined minimum aesthetic component score.

[0119] (4) A server according to any one of (1) to (3), wherein, in response to the received one or more images being from publicly available media corresponding to a set of pre-selected celebrities or trendsetters, the processing circuit system is further configured to: generate the character image database by defining or calculating a set of character images, each character image in the character image database having a typical human profile, the typical human profile being based on the physical characteristics and situational characteristics of the character in the one or more images, associating each pre-selected celebrity or trendsetter with a matching character image in the character image database, the association being based on the physical characteristics and situational characteristics of the pre-selected celebrity or trendsetter, and based on the aesthetic component score of the eyewear equipment of the pre-selected celebrity or trendsetter, associating the eyewear equipment with each character image in the character image set, the eyewear equipment associated with each character image corresponding to the one or more character image eyewear equipment profiles.

[0120] (5) A server according to any one of (1) to (4), wherein, in response to the received one or more images being one or more images of the user, the processing circuit system is further configured to: generate a profile for the user based on physical characteristics and situational characteristics of the user, associate the user with a matching character image in the character image database, the matching character image most closely matching the profile generated for the user based on the physical characteristics and situational characteristics of the user, and associate the user with a user eyewear equipment profile based on the matching character image, the user eyewear equipment profile being based on one or more character image eyewear equipment profiles of the matching character image.

[0121] (6) A server according to any one of (1) to (5), wherein the processing circuit system is further configured to: receive text corresponding to a usage description, train a vocabulary corresponding to the received text, and create a vector representation of the trained vocabulary.

[0122] (7) The server according to any one of (1) to (6), wherein the at least one aesthetic component includes: attributes of the one or more persons, situation attributes, and glasses attributes.

[0123] (8) The server of any one of (1) to (7), wherein the selected eyewear equipment is selected based on eyewear equipment available in a predetermined retailer catalog.

[0124] (9) A server according to any one of (1) to (8), wherein generating the character image database by calculating the character image set includes: extracting attributes from the received one or more images to construct the character image database based on publicly available media corresponding to the group of pre-selected celebrities or trendsetters.

[0125] (10) The server according to any one of (1) to (9), wherein a new character avatar database is generated at a predetermined interval corresponding to a known frequency of changes in fashion trends.

[0126] (11) A server according to any one of (1) to (10), wherein the aesthetic component score is based on a plurality of open source images from one or more of social media profiles, community reviews, and fashion magazines, the plurality of open source images corresponding to one or more pre-selected celebrities or trendsetters.

[0127] (12) A server according to any one of (1) to (11), wherein the processing circuit system for generating the aesthetic component score is further configured to: repeatedly receive the multiple open source images based on the frequency of trend changes in the fashion industry, and calculate a score for each of the multiple open source images based on a trend level, wherein the trend level is based on the number of times an item appears in the multiple open source images, the item including one or more of type, color, shape, product, pattern, brand and product category.

[0128] (13) A server according to any one of (1) to (12), wherein the score is weighed based on predetermined criteria including the time when the image in which the item appears was published, the time period over which all of the multiple images of the item were published, the person who provided the image, and feedback from others about the image, including social media engagement.

[0129] (14) The server of any one of (1) to (13), wherein the score is further weighted based on previous scores of the same aesthetic component from previous intervals.

[0130] (15) A method for selecting eyewear equipment, comprising: receiving one or more images by a processing circuit system, the one or more images including one or more representations of a person; applying a neural network by the processing circuit system to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images and generates an aesthetic component score for each of the one or more images; generating a user eyewear equipment profile for a user by the processing circuit system, the user being matched with a character image from a character image database, each character image in the character image database being linked to one or more character image eyewear equipment profiles, the one or more character image eyewear equipment profiles being based on the aesthetic component score; and selecting eyewear equipment for the user by the processing circuit system based on the generated user eyewear equipment profile.

[0131] (16) The method according to (15) further includes: summing the number of occurrences of each classified aesthetic component in the one or more images, wherein the number of occurrences is weighed according to predetermined criteria, the predetermined criteria including one or more of the following: the time when the image in which the item appears was published, the time period in which all the multiple images of the item were published, the person who provided the image, and feedback from others about the image, including social media engagement; and calculating an aesthetic component score for each of the one or more images based on the frequency with which each aesthetic component appears in the one or more images, wherein each aesthetic component is weighted based on one or more predetermined criteria and each aesthetic component in each of the one or more images is combined to calculate the aesthetic component statistics of the one or more images.

[0132] (17) The method according to (15) or (16) further includes: determining whether the user's eyewear equipment selection is based on the combination of all aesthetic components in the one or more images; in response to determining that the user's eyewear equipment selection is based on the attributes of all aesthetic components in the one or more images, outputting global statistics, wherein the global statistics is based on the aesthetic component score; determining whether the user's eyewear equipment selection is based on the eyewear aesthetic components in the one or more images; and in response to determining that the user's eyewear equipment selection is based on the eyewear aesthetic components in the one or more images, outputting eyewear statistics, wherein the eyewear statistics is based on the aesthetic component score, wherein the aesthetic component score is merged into the global statistics and the eyewear statistics only when the aesthetic component score is greater than a predetermined minimum aesthetic component score.

[0133] (18) The method according to any one of (15) to (17), in response to the received one or more images coming from publicly available media corresponding to a set of pre-selected celebrities or trendsetters, further includes: generating a character image database by defining or calculating a set of character images, each character image in the character image database having a typical human profile, the typical human profile being based on the physical characteristics and situational characteristics of the characters in the one or more images; associating each pre-selected celebrity or trendsetter with a matching character image in the character image database, the association being based on the physical characteristics and situational characteristics of the pre-selected celebrity or trendsetter; and associating eyewear equipment with each character image in the character image set based on the aesthetic component score of the eyewear equipment of the pre-selected celebrity or trendsetter, the eyewear equipment associated with each character image corresponding to the one or more character image eyewear equipment profiles.

[0134] (19) The method according to any one of (15) to (18), in response to the received one or more images being one or more images of the user, further includes: generating a profile for the user based on the physical characteristics and situational characteristics of the user; associating the user with the matching character image in the character image database, the matching character image most closely matching the profile generated for the user based on the physical characteristics and situational characteristics of the user; and associating the user with the user eyewear equipment profile based on the matching character image, the user eyewear equipment profile being based on the one or more character image eyewear equipment profiles of the matching character image.

[0135] (20) A non-transitory computer-readable storage medium having computer-readable instructions stored thereon, the computer-readable instructions, when executed by a computer, causing the computer to perform a method comprising: receiving one or more images, the one or more images comprising one or more representations of a person, applying a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, generating an aesthetic component score for each of the one or more images, generating a user eyewear configuration profile for a user, the user being matched to a persona from a persona database, each persona in the persona database being linked to one or more persona eyewear configuration profiles, the one or more persona eyewear configuration profiles being based on the aesthetic component score, and selecting eyewear configuration for the user based on the generated user eyewear configuration profile.

[0136] (21) The non-transitory computer-readable storage medium of (20), further comprising: summing the number of occurrences of each classified aesthetic component in the one or more images, wherein the number of occurrences is weighted according to predetermined criteria, the predetermined criteria comprising one or more of the following: the time when the image in which the item appears was published, the time period in which all the multiple images of the item were published, the person who provided the image, and feedback from others about the image, including social media engagement; and calculating an aesthetic component score for each of the one or more images based on the frequency with which each aesthetic component appears in the one or more images, wherein each aesthetic component is weighted based on the one or more predetermined criteria and each aesthetic component in each of the one or more images is combined to calculate aesthetic component statistics for the one or more images.

[0137] (22) The non-transitory computer-readable storage medium according to (20) or (21), further comprising: determining whether the user's eyewear equipment selection is based on a combination of all aesthetic components in the one or more images; in response to determining that the user's eyewear equipment selection is based on the attributes of all aesthetic components in the one or more images, outputting global statistics, wherein the global statistics are based on the aesthetic component scores; determining whether the user's eyewear equipment selection is based on the eyewear aesthetic components in the one or more images; and in response to determining that the user's eyewear equipment selection is based on the eyewear aesthetic components in the one or more images, outputting eyewear statistics, wherein the eyewear statistics are based on the aesthetic component scores, wherein the aesthetic component score is merged into the global statistics and the eyewear statistics only when the aesthetic component score is greater than a predetermined minimum aesthetic component score.

[0138] (23) According to any one of (20) to (22), the non-transitory computer-readable storage medium further comprises, in response to one or more images received from publicly available media corresponding to a set of pre-selected celebrities or trendsetters: generating a character image database by defining or calculating a set of character images, each character image in the character image database having a typical human profile, the typical human profile being based on the physical characteristics and situational characteristics of the characters in the one or more images; associating each pre-selected celebrity or trendsetter with a matching character image in the character image database, the association being based on the physical characteristics and situational characteristics of the pre-selected celebrity or trendsetter; and associating eyewear equipment with each character image in the character image set based on the aesthetic component score of the eyewear equipment of the pre-selected celebrity or trendsetter, the eyewear equipment associated with each character image corresponding to the one or more character image eyewear equipment profiles.

[0139] (24) According to the non-transitory computer-readable storage medium described in any one of (20) to (23), in response to the received one or more images being one or more images of the user, further comprising: generating a profile for the user based on the physical characteristics and situational characteristics of the user; associating the user with the matching character image in the character image database, the matching character image most closely matching the profile generated for the user based on the physical characteristics and situational characteristics of the user; and associating the user with the user eyewear equipment profile based on the matching character image, the user eyewear equipment profile being based on one or more character image eyewear equipment profiles of the matching character image.

Claims

1. A server for selecting eyewear equipment, comprising: processing circuitry configured to: receiving one or more images, the one or more images comprising one or more representations of a person, applying a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, the aesthetic component representing a particular item of aesthetic relevance, and generating an aesthetic component score for each of the one or more images based at least in part on a number of occurrences of each classified aesthetic component, generating a user eyewear outfit profile for the user by matching the user with a persona from a persona database, each persona in the persona database being linked to one or more persona eyewear outfit profiles, the user eyewear outfit profile being based on the one or more persona eyewear outfit profiles of the matched persona, the one or more persona eyewear outfit profiles being derived by associating eyewear outfits with personas based on the aesthetic component scores, and An eyewear outfit is selected for the user based on the generated user eyewear outfit profile.

2. The server according to claim 1, wherein: The processing circuitry is further configured to: summing the number of occurrences of each classified aesthetic component in the one or more images, wherein the number of occurrences is weighted according to predetermined criteria, the predetermined criteria comprising one or more of: the time when the image in which the aesthetic component appears was published, the time period during which all of the plurality of images including the aesthetic component were published, the person who provided the image, and feedback from others about the image, including social media engagement, and calculating the aesthetic component score for each of the one or more images based on a frequency of occurrence of each aesthetic component in the one or more images, Wherein, each aesthetic component is weighted based on one or more predetermined criteria, and each aesthetic component in each of the one or more images is combined to calculate aesthetic component statistics of the one or more images.

3. The server of claim 2, wherein: The processing circuitry is further configured to: determining whether the eyewear equipment selection of the user is based on a combination of all aesthetic components of the one or more images, In response to determining that the eyewear arrangement selection of the user is based on attributes of all aesthetic components of the one or more images, outputting global statistics, the global statistics being based on the aesthetic component scores, determining whether the eyewear outfit selection of the user is based on eyewear aesthetics in the one or more images, and In response to determining that the eyewear outfit selection of the user is based on an eyewear aesthetic component in the one or more images, outputting eyewear statistics, the eyewear statistics being based on the aesthetic component score, The aesthetic component score is merged into the global statistics and the glasses statistics only when the aesthetic component score is greater than a predetermined minimum aesthetic component score.

4. The server of claim 1, wherein: In response to receiving the one or more images from publicly available media corresponding to a set of pre-selected celebrities or trendsetters, the processing circuitry is further configured to: The character image database is generated by defining or calculating a set of character images, each character image in the character image database having a human-typical profile, the human-typical profile being based on physical features and situational features of the character in the one or more images, associating each pre-selected celebrity or trendsetter with a matching character profile in the character profile database, the association being based on physical and situational characteristics of the pre-selected celebrity or trendsetter, and Based on the aesthetic component score of the eyewear equipment of the pre-selected celebrity or trendsetter, eyewear equipment is associated with each character in the character set, and the eyewear equipment associated with each character corresponds to the one or more character eyewear equipment profiles.

5. The server of claim 4, wherein: In response to the received one or more images being one or more images of the user, the processing circuitry is further configured to: generating a profile for the user based on the physical and situational characteristics of the user, associating the user with a matching avatar in the avatar database that most closely matches the profile generated for the user based on the physical and situational characteristics of the user, and The user is associated with the user eyewear equipment profile based on the matching avatar.

6. A method for selecting eyewear equipment, comprising: receiving, by a processing circuit system, one or more images, the one or more images including one or more representations of a person; applying, by the processing circuit system, a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, the aesthetic component representing a particular item of aesthetic relevance, and generating an aesthetic component score for each of the one or more images based at least in part on a number of occurrences of each classified aesthetic component; generating, by the processing circuit system, a user eyewear outfit profile for the user by matching the user with a character avatar from a character avatar database, each character avatar in the character avatar database being linked to one or more character avatar eyewear outfit profiles, the user eyewear outfit profile being based on the one or more character avatar eyewear outfit profiles of the matched character avatar, the one or more character avatar eyewear outfit profiles being derived by associating eyewear outfits with character avatars based on the aesthetic component scores; and Eyewear equipment is selected, by the processing circuitry, for the user based on the generated user eyewear equipment profile.

7. The method of claim 6, further comprising: summing the number of occurrences of each classified aesthetic component in the one or more images, wherein the number of occurrences is weighted according to predetermined criteria, the predetermined criteria comprising one or more of: the time when the image in which the aesthetic component appears was published, the time period during which all of the plurality of images including the aesthetic component were published, the person who provided the image, and feedback from others about the image, including social media engagement; and calculating the aesthetic component score for each of the one or more images based on a frequency of occurrence of each aesthetic component in the one or more images, Wherein, each aesthetic component is weighted based on one or more predetermined criteria, and each aesthetic component in each of the one or more images is combined to calculate aesthetic component statistics of the one or more images.

8. The method of claim 7, further comprising: determining whether the eyewear equipment selection of the user is based on a combination of all aesthetic components of the one or more images; responsive to determining that the eyewear arrangement selection of the user is based on attributes of all aesthetic components of the one or more images, outputting global statistics, the global statistics being based on the aesthetic component scores; determining whether the eyewear outfit selection by the user is based on eyewear aesthetics in the one or more images; as well as In response to determining that the eyewear outfit selection of the user is based on an eyewear aesthetic component in the one or more images, outputting eyewear statistics, the eyewear statistics being based on the aesthetic component score, The aesthetic component score is merged into the global statistics and the glasses statistics only when the aesthetic component score is greater than a predetermined minimum aesthetic component score.

9. The method of claim 6, in response to receiving the one or more images from publicly available media corresponding to a set of pre-selected celebrities or trendsetters, further comprising: Generating the character image database by defining or calculating a set of character images, each character image in the character image database having a human-typical profile, the human-typical profile being based on physical features and situational features of the character in the one or more images; associating each pre-selected celebrity or trendsetter with a matching character profile in the character profile database, the association being based on physical and situational characteristics of the pre-selected celebrity or trendsetter; as well as Based on the aesthetic component score of the eyewear equipment of the pre-selected celebrity or trendsetter, eyewear equipment is associated with each character in the character set, and the eyewear equipment associated with each character corresponds to the one or more character eyewear equipment profiles.

10. The method of claim 9, in response to the received one or more images being one or more images of the user, further comprising: generating a profile for the user based on the physical and situational characteristics of the user; associating the user with the matching avatar in the avatar database, the matching avatar most closely matching the profile generated for the user based on the physical and situational characteristics of the user; as well as The user is associated with the user eyewear equipment profile based on the matching avatar.

11. A non-transitory computer-readable storage medium having computer-readable instructions stored thereon, the computer-readable instructions, when executed by a computer, causing the computer to perform a method comprising: receiving one or more images, the one or more images comprising one or more representations of a person, applying a neural network to the one or more images, wherein the neural network classifies at least one aesthetic component of each of the one or more images, the aesthetic component representing a particular item of aesthetic relevance, and generating an aesthetic component score for each of the one or more images based at least in part on a number of occurrences of each classified aesthetic component, generating a user eyewear outfit profile for the user by matching the user with a persona from a persona database, each persona in the persona database being linked to one or more persona eyewear outfit profiles, the user eyewear outfit profile being based on the one or more persona eyewear outfit profiles of the matched persona, the one or more persona eyewear outfit profiles being derived by associating eyewear outfits with personas based on the aesthetic component scores, and An eyewear outfit is selected for the user based on the generated user eyewear outfit profile.

12. The non-transitory computer readable storage medium of claim 11, further comprising: summing the number of occurrences of each classified aesthetic component in the one or more images, wherein the number of occurrences is weighted according to predetermined criteria, the predetermined criteria comprising one or more of: the time when the image in which the aesthetic component appears was published, the time period during which all of the plurality of images including the aesthetic component were published, the person who provided the image, and feedback from others about the image, including social media engagement; and calculating the aesthetic component score for each of the one or more images based on a frequency of occurrence of each aesthetic component in the one or more images, Wherein, each aesthetic component is weighted based on one or more predetermined criteria, and each aesthetic component in each of the one or more images is combined to calculate aesthetic component statistics of the one or more images.

13. The non-transitory computer readable storage medium of claim 12, further comprising: determining whether the eyewear equipment selection of the user is based on a combination of all aesthetic components of the one or more images; responsive to determining that the eyewear arrangement selection of the user is based on attributes of all aesthetic components of the one or more images, outputting global statistics, the global statistics being based on the aesthetic component scores; determining whether the eyewear outfit selection by the user is based on eyewear aesthetics in the one or more images; as well as In response to determining that the eyewear outfit selection of the user is based on an eyewear aesthetic component in the one or more images, outputting eyewear statistics, the eyewear statistics being based on the aesthetic component score, The aesthetic component score is merged into the global statistics and the glasses statistics only when the aesthetic component score is greater than a predetermined minimum aesthetic component score.

14. The non-transitory computer-readable storage medium of claim 11, in response to receiving the one or more images from publicly available media corresponding to a set of pre-selected celebrities or trendsetters, further comprising: Generating the character image database by defining or calculating a set of character images, each character image in the character image database having a human-typical profile, the human-typical profile being based on physical features and situational features of the character in the one or more images; associating each pre-selected celebrity or trendsetter with a matching character profile in the character profile database, the association being based on physical and situational characteristics of the pre-selected celebrity or trendsetter; as well as Based on the aesthetic component score of the eyewear equipment of the pre-selected celebrity or trendsetter, eyewear equipment is associated with each character in the character set, and the eyewear equipment associated with each character corresponds to the one or more character eyewear equipment profiles.

15. The non-transitory computer-readable storage medium of claim 14, in response to the received one or more images being one or more images of the user, further comprising: generating a profile for the user based on the physical and situational characteristics of the user; associating the user with the matching avatar in the avatar database, the matching avatar most closely matching the profile generated for the user based on the physical and situational characteristics of the user; as well as The user is associated with the user eyewear equipment profile based on the matching avatar.

Images