KR20220014543A - The measures and systems of online platform that connects both sides in a virtuous cycle through voluntary participation of users and developers. - Google Patents

Abstract

When a service user uploads a fashion product captured or saved using a mobile or PC camera, search results for similar products are provided through similar image search and, if shopping is desired, connected to an online platform. Collect and provide fashion big data to activate the initial platform. When anyone who wants to sell uploads their product image and inputs a sales route URL, these information can be provided to users as product candidates for similar product search services. When uploading product images to participate in sales, the results of product discrimination through pre-learned artificial intelligence are displayed according to the fashion category system. Items to be inputted directly are replaced with artificial intelligence and classifies products from the same perspective to provide users with accurate classification criteria. If the artificial intelligence determination result is determined to be misclassified, the resulting value is changed, and the misclassified value is used as a relearning material through transfer learning, contributing to the improvement of artificial intelligence accuracy. The present invention can advance the characteristics of easy switching between platform service users and sellers and the improvement of the core artificial intelligence of platform operators without incurring the cost of prior data pre-processing. The user can recommend the seller's product by registering the seller's product and registering the sales route. As the number of users and sellers increases, the registration of more voluntary products will replace big data collection. In addition, factors which are highly likely to be linked to purchases are analyzed to improve customer satisfaction with purchases and support effective marketing execution by sellers.

Description

{ The measures and systems of online platform that connects both sides in a virtuous cycle through voluntary participation of users and developers. }

According to an embodiment of the present invention, when a customer uploads a product picture, it finds the most similar product and connects it to an online shopping mall, and when a seller registers their product and stores an Internet sales address, voluntary participation of the user and the seller who connects the similar image It relates to the method and system of an online platform that is connected in a virtuous cycle.

With the development of artificial intelligence, it is possible to identify similar images and judge the similarity only with the image of fashion products, and with the development of big data collection and data processing technology, a service that collects and stores various fashion products It is possible.

Consumers can search for products more accurately when searching for products by image than by searching for products with keywords when shopping. Seller also wants to provide better shopping convenience to customers, but it requires a system to compare and connect data with images.

A service user can register a seller's product and promote sales on his behalf, or he can register his or her product as a seller at any time and propose a product to other service users on the platform.

When registering one's own registered product as a seller, AI classification technology already provides large, small, and small judgments that fit the fashion category classification system. In this case, transfer learning with artificial intelligence re-learning data is a method to increase the accuracy of artificial intelligence.

A platform service user can become a sales participant at any time, and a virtuous cycle link can be established to improve the accuracy of the artificial intelligence installed on the platform by checking for misclassification during the product registration process.

For the market and customer service, we provide initial services through data connection centered on various brands, create a synergy of mutual scale through an easy transition environment between users and sellers, and improve the quality of services by improving artificial intelligence in the platform in the process. .

According to the present invention, images are used instead of search words to enhance the shopping mall purchase convenience of service users, and the uploaded images can be connected to the online shopping mall by selecting a desired product while comparing a list of similar products recommended through a similarity determination. In addition, in order to recommend the seller's product, you can directly enter a photo and an online sales path to make a recommendation. To participate in this service, any seller can register as a similar image search product group just by registering a product, so that it can be used as a funnel for product sales. The registration procedure is minimized, and the result of determining the complicated product registration process through artificial intelligence is first provided. Through this connection platform, a platform that supports the connection between service users and sellers is provided, and for the AI evolution of the platform, the improvement process of voluntary participants' misclassification confirmation is systemized as a cycle of evolution. When response data is accumulated, personalized customer service is supported with customized recommendations such as trend suggestions, products preferred by peers, and brands.

The fashion big data information processing method, fashion product planning and product operation method and system according to the present invention for achieving the above object collect letters, numbers, and images collected through fashion-related online sales platforms and SNS /Storage unit: Fashion image data among the collected information is artificial intelligence image discrimination unit through artificial intelligence model learned by fashion classification system: Fashion obedience to collected character and numeric data, character for counting and ranking by item and classifier, Numerical data processing unit: Source data and processing data of images and alphanumeric characters, and analysis that constitutes a separator to apply filters for each user. Source data storage/management unit: Search fashion image photos, recommend similar products, and link to an online shopping mall Services that support the product and customer similar product inquiry/result, where customers recommend and register the seller’s product, recommended product Seller Link: Seller directly registers and manages fashion images and online sales URLs Product registration/management unit: customer inquiry similar to fashion image When a product is proposed, the selected product, a list connected to an Internet shopping mall, and a customer response data storage unit: Analyze the preferred brand or fashion type for each customer, and select the type of product selection by customer group such as age, gender, region, etc. It is characterized in that it includes a customer and seller response data analysis unit to analyze.

In addition, the letter, number, and image collection/storage unit stores information such as product name, price, actual price, sales quantity, stock information, and customer response (comments, retweets, recommendations) for each fashion product in a table form of a relational database structure. Character and number extraction/storage module ; Image extraction/storage module with a connector ( PRIMARY KEY ) as the file name for linking the images of fashion products with table-type data; may include

In addition, the artificial intelligence image determination unit, a supervised learning label management module for managing the target value obtained by classifying the determination target value for image determination as a label value suitable for fashion in a classification system; Faster RCNN model is used. Machine learning CNN learning module for data preprocessing and machine learning learning; Machine learning CNN re-learning module including re-learning and transfer learning of incorrect images to increase the accuracy of the target value; may include

In addition, the character and number data processing unit, a processing module for each classification system that processes the collected source information as a basic unit for analysis; A ranking module by collection unit that can be classified by brand unit, celebrity (power blog, power YouTuber, power SNS user) unit or grouped according to the same criteria; may include

In addition, the analysis source data storage/management unit includes: an image, character, and number connection module that integrates the results of the image determining unit and the character and numeric data processing unit into a connector ( PRIMARY KEY); Analysis item management module that processes connected data from the same point of view on a regular basis; a user authority module that allows access to target identifiers and item information for each user; may include

In addition, the customer similar product inquiry/result and recommended product seller connection unit searches fashion image photos, recommends similar products, and supports connection to an online shopping mall. a photo management module for uploading photos taken or saved by service users; The uploaded photo is a result inquiry module that judges similar products among pre-stored photos and delivers a list of similar products; Recommendation product registration seller connection module in which the user enters a product picture and online sales path to directly recommend the seller's product; may include

In addition, the seller product registration/management unit directly registers and manages fashion images and online sales URLs. A photo management module that the seller registers to expose the product to the service user; The registered photo provides the result value determined through artificial intelligence in the platform, and the seller confirms and confirms the photo confirmation module; photo registration history storage module that allows sellers to re-inquire and manage registered photos; may include

In addition, the customer and seller response data storage unit stores the selected product and a list connected to the Internet shopping mall when a product similar to the customer inquiry fashion image is proposed. Customer inquiry management module that manages product photos viewed by customers; A seller registration management module that manages the product photos registered by the seller; A seller connection record management module that manages the data that the customer inquires and connects to the online shopping mall; may include

In addition, the customer and seller response data analysis unit analyzes a brand or fashion type preferred by each customer, and analyzes a product selection type for each customer group such as age, gender, and region. Analysis module by customer characteristics that is aggregated, analyzed, and managed according to customer's gender and age; A seller analysis module that is classified, aggregated, analyzed, and managed according to the seller's connection diagram, obedience, and concept characteristics; Customer X seller cross-analysis module, which is divided, aggregated, analyzed and managed by cross-analyzing the purchase connection between customers and sellers or recommended customers and sellers; may include

It has innovation, advancement, and creativity in terms of a business model and system tool that will improve the efficiency of online shopping mall platform business by applying BIG.DATA and artificial intelligence to the fashion business.

A service user can easily register a product to promote the seller's product at any time, and can also register his or her own product as a seller. More product registrations create more users, and this interaction can create a larger number of participants.

During the seller's product registration process, the seller confirms the error in the AI classification as a confirmation process through the confirmation process, and this process can be expected to advance the artificial intelligence of the platform configuration through re-learning. Currently, the biggest challenge of artificial intelligence is the determination of the result value, the collection of data, and the labeling of the result.

Also service users. The response data to be accumulated according to the increase in the size of the seller becomes a customer individualized service by identifying the trend status, predicting the trend, suggesting the seller, and analyzing the product preferred by the peer group.

1 is a block diagram of a method and system of an online platform that is connected in a virtuous cycle through voluntary participation of users and sellers according to the present invention.
2 is a block diagram of a character, number, and image collection/storage unit according to the present invention.
3 is a block diagram of an artificial intelligence image discrimination unit according to the present invention.
4 is a block diagram of a character and numeric data processing unit according to the present invention.
5 is a block diagram of an analysis source data storage/management unit according to the present invention.
6 is a block diagram of a customer similar product inquiry/result and recommended product seller connection unit according to the present invention.
7 is a configuration diagram of a seller product registration/management unit in the present invention.
8 is a customer and seller response data storage unit according to the present invention.
9 is a customer and seller response data analysis unit according to the present invention.
10 is a conceptual diagram of a method and system of a virtuous cycle connection online platform according to the present invention.
11 is an exemplary view of the user_customer similar product inquiry/recommendation service unit visualization according to the present invention.
Fig. 12 is an example of the visualization of the seller product registration/management unit according to the present invention.
13 is a diagram of a fashion category classification system according to the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. First of all, it should be noted that the same elements in the drawings are denoted by the same reference numerals whenever possible.

1 is a block diagram of a method and system providing apparatus of an online platform that is connected in a virtuous cycle through voluntary participation of users and sellers according to an embodiment of the present invention.

The method and system of an online platform that is connected in a virtuous cycle through the voluntary participation of users and sellers according to the present invention, as shown in FIG. , character, numeric data processing unit 300 , analysis source data storage/management unit 400 , customer service unit 500 , seller management unit 600 , customer, seller reaction data storage unit 700 , customer, seller reaction data and an analysis unit 800 .

Visualizing such a service is the same as FIG. 10, a conceptual diagram of the method and system of the virtuous cycle connection online platform.

2 is a block diagram of a character, number, and image collection/storage unit according to the present invention.

As shown in FIG. 2 , the character, number, and image collection/storage unit according to the present invention includes a character/number extraction/storage module 110 and an image extraction/storage module 120 .

The letter, number, and image collection/storage unit collects and stores big data by classifying it into images, letters, and numbers according to its characteristics.

3 is a block diagram of an artificial intelligence image discrimination unit according to the present invention.

As shown in FIG. 3, the artificial intelligence image discrimination unit according to the present invention includes a supervised learning label management module 210, a machine learning image recognition convolutional neural network (CNN) learning module 220, a machine learning convolutional network (CNN) Neural Network ) includes a re-learning module 230 .

The supervised learning label management module 210 is used as a label value (target value) in supervised learning of machine learning for artificial intelligence image discrimination learning. When an image is input, supervised learning is performed to judge it according to the classification system, and this label value is arranged according to a standard suitable for fashion.

The label value is the same as the large-medium-small classification system, which is a category system, and the system is the same as the pattern 8 fashion category classification system.

The machine learning CNN (Convolutional Neural Network) learning module 220 learns by dividing the discrimination model through artificial intelligence into 1 large, 4 medium, and 13 small, respectively. In order to judge a single image, the image is classified into large-medium-small steps. In the case of outer classification in the large classification, COAT, JACKET in the middle classification stage

It is classified as one of ,JUMPER, JUMPER_ PARKA, and VEST. If it is classified as COAT, it is classified among the sub-categories in COAT. Ultimately, the goal is to increase the image accuracy through artificial intelligence while narrowing the discrimination according to categories.

The machine learning CNN re-learning module 230 uses the misclassified image again as transfer learning material of artificial intelligence when the judgment result is different when a person judges.

The image determined by misclassification through the photo confirmation module 620 of the configuration diagram of the seller management unit 600 is used as high-learning data of the machine learning CNN re-learning module 230 .

4 is a block diagram of a character and numeric data processing unit according to the present invention.

As shown in FIG. 4, the character and numeric data processing unit according to the present invention includes a processing module 310 for each classification system and a ranking module 320 for each collection unit.

The processing module 310 for each classification system is a fashion category classification system (large-medium-small), and for each classification, the total and size of the number of styles, sales quantity, stock quantity, customer response information (likes, comments, retweets), etc. Pre-configure the ranking according to

The ranking module 320 by collection unit collects data according to the total and size of the number of styles, sales quantity, stock quantity, customer response information (likes, comments, retweets), etc. by brand, seller, concept, and arbitrary grouping. Pre-configure the ranking.

5 is a block diagram of an analysis source data storage/management unit according to the present invention.

As shown in FIG. 5 , the analysis source data storage/management unit according to the present invention includes an image, character, and number connection module 410 , an analysis item management module 420 , and a user authority module 430 .

The image, character, and number connection module 410 connects the result obtained through the image determining unit 200 and the character, number, and data processing unit 300 according to the nature of the data with a connector ( PRIMARY KEY ) to integrate data do.

The analysis item management module 420 determines the unit of analysis to fit the service and system users. In the case of classifying only the items of a specific product according to the user's needs, it is configured to analyze the core competition scope only for a specific brand and specific obedience.

The user authority module 430 is configured so that only the items the user wants to use among the types of services to be provided are provided as service items. The analysis item management module 420 is managed for each user. At this time, the authority and the period of authority are configured to be managed together.

6 is a customer similar product inquiry/result recommendation product seller connection unit according to the present invention.

As shown in FIG. 6 , the customer service unit according to the present invention includes a photo management module 510 , a result inquiry module 520 , and a recommended product registration seller connection module 530 .

This service is shown in Figure 11. It is the same as the visualization example diagram of the user_customer similar product inquiry/recommendation service department.

The photo management module 510 is. When a customer uploads a registered photo to find a similar image through mobile or PC, the image is embedded with the output result of the artificial intelligence layer before final label determination through an image recognition convolutional neural network (CNN: Convolutional Neural Network). embedding) includes a module for converting vector values.

The result inquiry module 520 compares the product image collected through big data with a value converted into an embedding vector value through a convolutional neural network. At this time, a result inquiry module is included that finds the most similar result to the customer image input through the Nearest Neighbor search technique and provides the result value to the customer.

The recommended product registration seller connection module 530 supports a customer recommendation function in which a user directly inputs an image and an online sales path of the seller's product to advertise on behalf of the seller. Anyone can participate in the promotion of the seller by the user, and it includes an interconnection module between the user and the seller.

7 is a configuration diagram of a seller product registration/management unit in the present invention.

As shown in FIG. 7 , the fashion seller management unit according to the present invention includes a photo management module 610 , a photo confirmation module 620 , and a photo registration history storage module 630 . A service user can become a seller at any time. If there is a virtuous cycle of mutual conversion, platform users can quickly expand. When service users and sellers circulate to induce scale growth, products are diversified and provided, and the number of users increases.

Supervised learning of artificial intelligence requires the target value (label ) to have an accurate image. As the learning data for this is directly collected and adjusted through the participants, the process of pre-data processing for machine learning learning can be shortened, and as the scale increases, it creates an opportunity to improve the accuracy of artificial intelligence.

This service is shown in Figure 12. It is the same as the visualization example of the seller product registration/management unit.

The photo management module 610 is composed of a photo management module in which the seller directly performs the steps of uploading an image, determining a fashion category value of a product, and inputting a URL for selling an Internet product in order to register a product for sale.

The photo confirmation module 620, the registered photo is a fashion category (large-large-small) classification of the product for the convenience of seller registration, and gives a result value through artificial intelligence learned by machine learning, and the corresponding Edit it within the category. 13 according to the present invention. Classify according to the fashion category classification system diagram. When the result is considered to be misclassified, the photo confirmation module is directly changed to a suitable fashion category and then confirmed.

The photo registration history storage module 630 consists of a photo registration history storage module that manages photos of fashion products registered by the seller and records of registration information.

8 is a block diagram of a customer and seller response data storage unit according to the present invention.

As shown in FIG. 8 , the customer and seller response data storage unit according to the present invention includes a customer inquiry management module 710 , a seller registration management module 720 , and a seller connection record management module 730 .

The customer inquiry management module 710 includes a customer inquiry management module that manages a connection history including customer characteristics (gender, age).

The seller registration management module 720 includes a seller registration management module that manages sellers participating in product registration for the purpose of sales according to fashion obedience, concept, and participation.

The seller connection record management module 730 is configured as a seller connection record management module that manages the history of connecting to an Internet shopping mall after inquiring a similar image of a customer.

9 is a configuration diagram of a customer and seller response data analysis unit according to the present invention.

As shown in FIG. 8 , the customer and seller response data analysis unit according to the present invention includes an analysis module for each customer characteristic 810 , a seller analysis module 820 , and a customer X seller cross-analysis module 830 .

Preference analysis according to customer's gender and age group data, brand, and concept provides analysis results for trend status check, trend prediction, peer group preference product recommendation, and suitable seller connection service.

The analysis module for each customer characteristic 810 includes an analysis module for each customer characteristic that analyzes the type of selection according to the customer characteristics (gender, age).

The seller analysis module 820 includes a seller analysis module that analyzes the type of product, fashion obedience, concept, and user connection level registered for each seller.

The customer X seller cross-analysis module 830 consists of a customer X seller cross-analysis module that analyzes the degree of response for each customer and seller.

100: character, number, image collection/storage unit 110: character, number extraction/storage module
120: image extraction / storage module 200: artificial intelligence image identification unit
210: supervised learning label management module 220: machine learning CNN learning module
230: machine learning CNN re-learning module 300: character, numeric data processing unit
310: Processing module by classification system 320: Ranking module by collection unit
400: Analysis source data storage/management unit 410: Image, character, number connection module
420: analysis item management module 430: user authority module
500: Customer similar product inquiry/result, recommended product seller link
510: photo management module 520: result inquiry module
530: Recommended product registration Seller connection module 600: Seller product registration/management unit
610: photo management module 620: photo confirmation module
630: photo registration history storage module 700: customer, seller response data storage unit
710: customer inquiry management module 720: seller registration management unit
730: seller connection record management unit 800: customer, seller response data analysis unit
810: analysis module by customer characteristics 820: seller analysis module
830:CustomerXSeller cross-analysis module

Claims (9)

A character, number, and image collection/storage unit that classifies fashion big data according to the nature of the data; An artificial intelligence image discrimination unit that determines an image classification value through identification of images among the collected data; Text mining technique or arithmetic operation of the collected data, character and numeric data processing unit; Combined storage of collected data as a connector ( PRIMARY KEY ) and analysis source data storage/management unit reflecting the usage environment; A service that recommends similar products by inquiring fashion image photos and supports connection to an online shopping mall, and a customer similar product inquiry/result, recommended product seller connection unit for customers to recommend and register the seller's products; Merchant product registration/management unit that directly registers and manages fashion images and online sales URLs; Customer and seller response data storage unit that stores the selected product when a product similar to the customer inquiry fashion image is proposed and a list linked to an Internet shopping mall; Connecting a virtuous cycle through voluntary participation of users and sellers, characterized in that it includes a customer and seller response data analysis unit that analyzes the brand or fashion type preferred by each customer, and analyzes the type of product selection by customer group such as age, gender, region, etc. It is the method and apparatus of the online platform method and system.
The text, number, and image collection/storage unit 100 extracts/stores text and numbers such as product name, product-related information and sales volume, customer response information (comments, number of comments, …) on a fashion e-commerce platform or fashion-related SNS The module 110 includes an image extraction/storage module 120 for extracting and storing image information of the product.
The artificial intelligence image determination unit 200 includes a supervised learning label management module 210 , a machine learning CNN learning module 220 , and a machine learning CNN re-learning module 230 . Supervised learning label management module (210, ) that creates classification criteria to match the target value (label) to the fashion category classification system in supervised learning and manages the flow of large-medium-small (210, ) collected images taken from Machine learning CNN learning module 220, which performs the actual machine learning learning process, including pre-processing before machine learning learning, and machine learning CNN re-learning including the process of re-learning to increase the accuracy through TEST of the learning result module 230 .
The character and number data processing unit 300 includes a processing module 310 for each classification system and a ranking module 320 for each collection unit. The collected letter and number information related to fashion is based on the fashion classification (large-medium-small) system and data source. Pre-aggregation is required to quickly perform analysis by brand unit. A processing module for each classification system that aggregates and processes data in advance by classification, source, and unit (310), and also pre-aggregates priority, weight, and cumulative weight by size and number of product styles according to customer response. and a ranking module 320 for each collection unit.
The analysis source data storage/management unit 400 includes an image, text, and number connection module 410 , an analysis item management module 420 , and a user authority module 430 . The image, letter, and number connection module 410 that combines the data added through the letter and number processing unit and the classified and coded data for each image by the image processing unit as a PRIMARY KEY as a connector ( PRIMARY KEY ). It is necessary to focus on the item unit only. An analysis item management module 420 that can separately analyze and aggregate the brand or product unit that the user wants, manages the authority for each user's service use item so that the customer can select and provide the service desired by the customer among the services to be provided through data includes a user rights module 430 .
It includes the customer similar product inquiry/result, recommended product seller connection unit 500 , photo management module 510 , result inquiry module 520 , and recommended product registration seller connection module 530 . A photo management module (510) that stores and manages photos registered by the customer to use the similar image search service, and a result inquiry module (520) that manages similar image results suitable for the searched image and manages and stores the image selected by the customer (520) , a recommended product registration seller connection module 530 in which the user registers and manages photos and online sales addresses for seller product recommendation.
The seller product registration/management unit 600 includes a photo management module 610 , a photo confirmation module 620 , and a photo registration history storage module 630 . The photo management module 610 for registering products for sale, the registered photos automatically classify fashion categories (large-greater-small) through artificial intelligence, and the seller confirms and finally confirms. and a photo confirmation module 620 for directly changing to a suitable fashion category and confirming the result when it is considered to be misclassified, and a photo registration history storage module 630 for managing a history of photos and registration information of registered fashion products.
claim 8
The customer and seller response data storage unit 700 includes a customer inquiry management module 710 , a seller registration management module 720 , and a seller connection record management module 730 . Customer inquiry management module 710 that manages connection history including customer characteristics (gender, age), seller registration management module 720 that manages sellers who have participated in product registration for the purpose of sales, after inquiring similar images of customers and a seller connection record management module 730 that manages the history of connecting to the Internet shopping mall.
The customer and seller response data analysis unit 800 includes an analysis module 810 by customer characteristics, a seller analysis module 820 , and a customer X seller cross-analysis module 830 . Analysis module by customer characteristics (810) that analyzes the type of selection according to customer characteristics (gender, age), seller analysis module (820) that analyzes the type of product registered for each seller, and the degree of connection, degree of response by customer and seller and a customer x seller cross-analysis module 830 to analyze .

Images