KR102495868B1 - Fashion-related customized perfume recommendation system using ai - Google Patents

Abstract

The present invention relates to a system for recommending fashion-related customized perfume by using artificial intelligence (AI) and, more specifically, to a system for recommending fashion-related customized perfume by using AI, which utilizes AI technology to classify fashion styles of a user and select fragrance types to recommend fashion-related customized perfume. The system for recommending fashion-related customized perfume by using AI according to the present invention comprises: an input unit receiving fashion coordination images from a user terminal through a perfume recommendation service; a classification unit classifying the user's fashion styles from a plurality of preset styles, based on characteristic information of item objects detected for each part from the fashion coordination images; a selection unit using a pre-learned AI-based fashion-perfume linkage model to select the user's fragrance types corresponding to the user's fashion styles and the characteristic information among a plurality of previously collected fragrance styles; a search unit searching for a plurality of perfume products containing the user's fragrance styles; and a recommendation unit generating a customized perfume list page, based on the product information of each of the plurality of perfume products and whether or not the same kinds of perfumes having the user's fragrance styles are present to recommend the same to the user terminal. Accordingly, the user's fashion styles and the user's fragrance styles are more accurately derived and linked to each other to provide objective and visible decision standards for perfume products.

Description

Fashion-linked personalized perfume recommendation system using AI {FASHION-RELATED CUSTOMIZED PERFUME RECOMMENDATION SYSTEM USING AI}

The present invention relates to a fashion-linked perfume recommendation system using AI, and more specifically, to recommend a customized perfume product linked to fashion by classifying a user's fashion style and selecting a scent series using artificial intelligence technology. It is about a customized perfume recommendation system linked to fashion using artificial intelligence (AI).

Perfume is regarded by people as part of fashion and as a means of self-expression. However, in the case of perfume, it is not easy for users (consumers) to find the scent they want unless they personally try the perfume in a store.

On the other hand, with the development of technology, various online shopping malls have been created on the Internet that sell (or trade, deliver, etc.) perfume so that users can purchase perfume (perfume products) even if they do not visit the store directly. More diverse types of perfume products than offline are being actively traded.

However, when purchasing a perfume product sold in an online shopping mall, it is not possible to directly try the product or check the actual size, so when purchasing a perfume product simply by looking at the information on the perfume product provided on the shopping mall, actually receive and view the product. Cases in which the purchased product (perfume product) is different from the scent the buyer thought or the size is different from what the buyer thought (for example, the size of the perfume bottle is too large or small than expected) occur frequently, resulting in The user's satisfaction with the purchased perfume product is not high.

In this case, the user takes actions such as returning, returning, or exchanging the perfume product, which causes time and financial damage to the shopping mall operator, product producer, and buyer.

(0001) Korean Patent Publication No. 10-2017-0096911

The present invention is to solve the above problems, and an object of the present invention is to provide a fashion-linked customized perfume recommendation system using AI that can derive a user fashion style and a user scent series according to a coordinating image.

In addition, based on the perfume information of the perfume products searched according to the user fragrance series, it is to provide a fashion-linked customized perfume recommendation system using AI that can recommend customized perfume products in the form of a web page.

The above and other objects and advantages of the present invention will become apparent from the following description of preferred embodiments.

In order to achieve the above object, according to an embodiment of the present invention, a fashion-linked customized perfume recommendation system using AI includes an input unit that receives a coordinating image through a perfume recommendation service from a user terminal, and a part-by-part detection from the coordinating image. Based on the feature information of the item object, using a classification unit that classifies user fashion styles from a plurality of preset styles and a pre-learned artificial intelligence-based fashion scent linkage model, among a plurality of pre-collected fragrance series A selection unit that selects a user fragrance series corresponding to the user fashion style and the feature information, a search unit that searches for a plurality of perfume products containing the user fragrance series, and each product information of the plurality of perfume products and the user fragrance series and a recommendation unit for generating a customized perfume list page and recommending it to the user terminal based on whether or not the liver is the same type.

In an embodiment, the input unit draws the coordinating image based on the preferred image for each part selected by the user as sample images are provided through the perfume recommendation service.

In an embodiment, the classification unit uses a feature detection unit that detects feature information through object recognition analysis for the item object detected for each part from the coordinating image, and a pre-learned artificial intelligence-based style classification model, It includes a similarity output unit outputting a similarity of information for each style and a style registration unit classifying and registering at least one style having a similarity of a predetermined value or more among the plurality of styles as a user style.

In an embodiment, the similarity output unit learns the style classification model in advance through machine learning that takes a plurality of pre-collected feature information as an input and outputs a plurality of pre-collected style categories, and outputs the feature information. An output value derived by applying the style classification model is determined as a similarity to the feature information.

In an embodiment, the selector includes a collection unit that collects a plurality of first perfume series mapped by style through fashion and fragrance experts and a plurality of second scent series mapped by feature information; A learning unit that learns an artificial intelligence-based fashion scent linkage model through machine learning that takes styles and feature information as inputs and outputs the plurality of first and second fragrance series mapped according to the style and feature information. , an accuracy determination unit for determining the accuracy of an output value for each perfume series output by applying the user fashion style and the feature information to the fashion scent link model, and at least one scent series having the accuracy of a certain value or more as a user fragrance series. Including the register of fragrance families that are selected and registered.

In an embodiment, the learning unit adjusts a weight for the fashion scent association model based on the number of purchase history and the number of sample requests of the customized perfume list page generated according to the user scent sequence.

In an embodiment, based on the psychological state of the user confirmed through the perfume recommendation service from the user terminal, a ranking unit for automatically determining a list ranking of the plurality of perfume products according to scent sequence intensity; Includes a listing unit that lists perfume products of a predetermined rank or higher among perfume products and creates a customizable perfume list page that is payable, and a page providing unit that provides the customized perfume list page to the user terminal through the perfume recommendation service. do.

In an embodiment, when the user's psychological state is confirmed to be in an unstable state, the ranking unit determines the list ranking of the plurality of perfume products in the order of fragrance sequence intensity, and the user's psychological state is in a stable state. If it is confirmed, the list ranking of the plurality of perfume products is determined in the order of low perfume intensity.

In an embodiment, the input unit searches for a fashion image for each part corresponding to text information received from the user terminal through the perfume recommendation service, and draws a coordinating image using the fashion image for each part.

In an embodiment, the recommendation unit adjusts a preset perfume duration time for each perfume product in inverse proportion to body temperature information received from the user terminal through the perfume recommendation service, and sets the reuse time according to the usage time of the perfume in the form of an alarm. guide

According to an embodiment of the present invention, by more accurately deriving and linking a user's fashion style and a user's perfume series, it is possible to provide an objective and visible criterion for determining perfume products.

In addition, by recommending customized perfume products in the form of a web page based on the perfume information of the perfume products searched for according to the user fragrance series, it is possible to support the purchase of the desired perfume product without the user's experience.

1 is a diagram schematically showing a fashion-linked customized perfume recommendation system 1000 using AI according to an embodiment of the present invention.
2(A) and 2(B) are embodiments of a perfume recommendation service.
3(A) and 3(B) are examples for explaining each operation of the classification unit 200, FIG. 3(C) is an example for explaining the operation of the selection unit 300, and FIG. 3(D) ) is an example for explaining the operation of the search unit 400.
FIG. 4A is a block diagram of the classification unit 200 of FIG. 1 according to one embodiment, and FIG. 4B is a block diagram of the classification unit 200 of FIG. 1 according to another embodiment.
FIG. 5 is a block diagram of the selector 300 of FIG. 1 according to an embodiment.
FIG. 6 is a block diagram of the recommendation unit 500 of FIG. 1 according to an embodiment.

Hereinafter, the present invention will be described in detail with reference to embodiments and drawings of the present invention. These examples are only presented as examples to explain the present invention in more detail, and it will be apparent to those skilled in the art that the scope of the present invention is not limited by these examples. .

In addition, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of skill in the art to which this invention belongs, and in case of conflict, this specification including definitions of will take precedence.

In order to clearly explain the proposed invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification. And, when a certain component is said to "include", this means that it may further include other components without excluding other components unless otherwise stated. Also, a “unit” described in the specification means one unit or block that performs a specific function.

In each step, the identification code (first, second, etc.) is used for convenience of description, and the identification code does not describe the order of each step, and each step does not clearly describe a specific order in context. It may be performed differently from the order specified above. That is, each step may be performed in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

1 is a diagram schematically showing a fashion-linked customized perfume recommendation system 1000 using AI according to an embodiment of the present invention, and FIGS. 2(A) and 2(B) are embodiments of a perfume recommendation service. 3(A) and 3(B) are examples for explaining each operation of the classification unit 200, FIG. 3(C) is an example for explaining the operation of the selection unit 300, and FIG. 3( D) is an example for explaining the operation of the search unit 400.

Referring to FIGS. 1 to 3 (D), the fashion-linked personalized perfume recommendation system 1000 using AI includes an input unit 100, a classification unit 200, an extraction unit 300, and a search unit 400. and a recommendation unit 500 .

First, the input unit 100 may receive a coordinating image from the user terminal 10 through a perfume recommendation service and register it in the storage DB 50 .

Here, the coordinating image may be an image including other users or clothes, shoes, and fashion items worn by the user.

At this time, the perfume recommendation service is a plurality of user terminals (e.g., 10_1 to 10_N) connected from the input unit 100 through a web server (not shown), as shown in FIGS. 2(A) and 2(B). It may be an app or program provided and installed on the .

This perfume recommendation service can be implemented as a graphic user interface that is input and selected through at least one method of image capture, sample image-based touch selection input, and image upload as customer information registration is completed.

According to an embodiment, the input unit 100 may draw a coordinating image based on the preferred image for each part selected by the user through the user terminal 10 as sample images are provided through the perfume recommendation service.

According to another embodiment, the input unit 100 may search for a fashion image for each part corresponding to text information received from the user terminal 10 through a perfume recommendation service. At this time, the input unit 100 may draw a coordinating image using the fashion image for each part.

According to another embodiment, the input unit 100 may apply a preset correction filter according to age, gender, and occupation to a coordinating image based on user information received from the user terminal 10 through the perfume recommendation service. there is.

According to another embodiment, the input unit 100 selectively selects a preset correction filter for a coordinating image based on a preset change value of a user's bio-signal received from a bio-signal measurer (not shown) through a perfume recommendation service. can be applied.

Here, the bio-signal measurer (not shown) is electrically connected to the user terminal 10, and measures at least one of the user's electrocardiogram, body temperature, blood pressure, pulse rate, and blood sugar level, and transmits it to the input unit 100 through a perfume recommendation service. there is. In this case, the preset correction filter may be an effect for applying any one of brightness, white balance, contrast, hue, color, sharpness, shading, and contrast to the image.

For example, if the user's bio-signal is greater than or equal to a preset change value, the input unit 100 applies a preset correction filter to the coordinating image, and if the user's bio-signal is less than the preset change value, the input unit 100 applies the coordinating image A preset correction filter may not be applied to .

Next, the classification unit 200 selects a plurality of styles pre-collected in the storage DB 50 based on the feature information of the item object detected for each part from the coordinating image received through the input unit 100 or drawn. User fashion styles can be classified.

Here, the item object may mean any one object of top, bottom, outerwear, dress, and shoes. In this case, the feature information may include information on any one of color, detail, print, material, fit, length, type, sleeve, neckline, and pattern.

In addition, a plurality of previously collected styles include the first category including traditional, mannish, feminine, ethnic, contemporary, natural, genderfluid, sporty, subculture and casual, American casual, casual, chic, formal, dandy, girlish, The second category includes Golf, Retro, Romantic, Sports, Street and Beauty, and Classic, Preppy, Manish, Tomboy, Feminine, Romantic, Sexy, Hippie, Western, Oriental, Modern, Sophisticated, Avant-gardo, Continental, A third category including resort, genderless, sporty, retro, kitsch/kidult, hip-hop, punk, military, and street may be included.

Specifically, as shown in FIG. 3(A), the classification unit 200 performs object recognition analysis on item objects detected for each part from the coordinating image, for example, shoes, slacks, goodies, leather jackets, and shirts. Characteristic information is detected for each part, and as shown in FIG. 3(B), the user fashion style is based on at least one style corresponding to the feature information for each part among a plurality of styles pre-collected in the storage DB 50. can be classified.

Next, as shown in FIG. 3(C), the selection unit 300 uses the pre-learned artificial intelligence-based fashion scent association model among a plurality of fragrance series pre-collected in the storage DB 50. It is possible to select a user-oriented series corresponding to user fashion style and feature information.

Here, the plurality of fragrance series may include musk, aromatic, woody, citrus, oriental, metallic, honey, green, leathery, marine, powdery and floral.

At this time, as shown in FIG. 3(C), the pre-learned artificial intelligence-based fashion trend linkage model takes a plurality of pre-collected styles and feature information as inputs, and pluralities mapped according to the style and feature information. It may be an artificial neural network (ANN) learned through machine learning with the first and second flavor series of as outputs. That is, the artificial intelligence-based fashion trend linkage model may be any one of an artificial neural network, a support vector machine (SVM), a decision tree, and a random forest.

For example, an artificial neural network is a statistical learning algorithm that is mainly used in deep learning and is inspired by neural networks in machine learning and biology, and may be a convolutional neural network including a feature extraction neural network and a classification neural network. At this time, the convolutional neural network is a type of deep feed-forward artificial neural network used to analyze visual images, and can be divided into processes of extracting features of images and classifying them, and extracting and extracting features of specific images. Images can be recognized based on their features.

Specifically, the feature extraction neural network proceeds by sequentially stacking the convolution layer and the pooling layer on the input signal. The convolution layer contains convolution operations, convolution filters, and activation functions. The calculation of the convolution filter is adjusted according to the matrix size of the target input, but a 9X9 matrix is generally used. The activation function may generally use a ReLU function, a sigmoid function, a tanh function, etc., but is not limited thereto. The pooling layer is a layer that reduces the matrix size of the input, and uses a method of extracting a representative value by grouping pixels in a specific area. In general, the average value or the maximum value is often used for the calculation of the pooling layer, but is not limited thereto. The operation is performed using a square matrix, and a 9X9 matrix is generally used. The convolutional layer and the pooling layer are alternately repeated until the corresponding input is sufficiently small while maintaining the difference.

Also, a classification neural network has a hidden layer and an output layer. Such a classification neural network generally includes 5 or more hidden layers, and the number of nodes in each hidden layer can be designated as 80, but it is possible to set more than 80 in some cases. The activation function of the hidden layer uses a ReLU function, a sigmoid function, a tanh function, and the like, but is not limited thereto.

Next, as shown in FIG. 3(D), the search unit 400 may search through the storage DB 50 for a plurality of perfume products containing user fragrance series registered through the selection unit 300. there is.

Depending on the embodiment, the search unit 400 may search for a plurality of perfume products through a keyword web search for at least one fragrance series selected through the selection unit 300.

Next, the recommender 500 creates a customized perfume list page based on the product information of the plurality of perfume products searched through the search unit 400 and whether or not the user perfume series is the same type, and creates a perfume recommendation service. It can be recommended to the user terminal 10 through.

Here, the product information may include brand, product type, product characteristics, fragrance series type, fragrance series intensity, price, capacity, image, use time and use season.

According to an embodiment, the recommendation unit 500 may provide a sorting function through a perfume recommendation service so as to rearrange the order listed on the customized perfume list page based on product information.

According to another embodiment, the recommendation unit 500 may capture facial images at regular intervals through a perfume recommendation service using a camera (not shown) provided in the user terminal 10 .

At this time, the recommendation unit 500 may automatically output usage review information for each perfume product listed on the customized perfume list page based on a preset facial expression object detected from the face image.

For example, the preset expression object may include an expressionless expression object, an unstable expression object, a suspicious expression object, a confused expression object, a scary expression object, an urgent expression object, and the like.

According to another embodiment, the recommender 500 may determine whether or not to be in a drowsy state based on a predetermined eyelid-shaped object detected from a face image, and may recommend a perfume product for preventing drowsiness by integrating it into a customized perfume list page.

Here, the perfume product for preventing drowsiness may refer to a perfume product containing herbal and green tea flavors.

According to another embodiment, the recommender 500 uses the MBTI personality test program provided through the perfume recommendation service from the user terminal 10 to check the user's personality test result, and according to the personality test result, the maximum number of Perfume products sold as can be integrated into a customized perfume list page and recommended.

According to another embodiment, the recommendation unit 500 may filter some perfume products from the customized perfume list page based on feedback information extracted from user review information for each perfume product.

According to another embodiment, the recommendation unit 500 determines the intensity of each scent sequence corresponding to the external activity and the internal activity based on the noise intensity detected through a microphone (not shown) provided in the user terminal 10. It can be set to the user's preferred strength.

According to another embodiment, the recommendation unit 500 adjusts the preset perfume duration for each perfume product in inverse proportion to the body temperature information received from the user terminal 10 through the perfume recommendation service, and the reuse time according to the perfume use time. can be guided through a perfume recommendation service in the form of an alarm.

Hereinafter, the configuration of the present invention and its effects will be described in more detail through specific examples and comparative examples. However, these examples are for explaining the present invention in more detail, and the scope of the present invention is not limited to these examples.

FIG. 4A is a block diagram of the classification unit 200 of FIG. 1 according to one embodiment, and FIG. 4B is a block diagram of the classification unit 200 of FIG. 1 according to another embodiment.

Referring to FIGS. 1 , 4A and 4B , the classification unit 200 may include a feature detection unit 210 , a similarity output unit 220 and a style registration unit 230 .

First, the feature detector 210 may detect feature information through object recognition analysis on item objects detected for each part from a coordinating image.

Next, the similarity output unit 220 may output the similarity of feature information for each style using the pre-learned artificial intelligence-based style classification model.

According to an embodiment, the similarity output unit 220 learns an artificial intelligence-based style classification model in advance through machine learning that takes a plurality of pre-collected feature information as input and outputs a plurality of pre-collected style categories. can do.

At this time, the similarity output unit 220 may output an output value for each style derived by applying the feature information detected through the feature detector 210 through a style classification model as a similarity to the feature information.

Next, the style registration unit 230 may classify at least one style having a similarity to feature information from a plurality of styles as a user style, integrate it into user information, and register it in the storage DB 50 .

FIG. 5 is a block diagram of the selector 300 of FIG. 1 according to an embodiment.

Referring to FIGS. 1 and 5 , the selection unit 300 may include a collection unit 310, a learning unit 320, an accuracy determination unit 330, and a fragrance series registration unit 340.

First, the collection unit 310 may collect a plurality of first scent series mapped by style and a plurality of second scent series mapped by feature information through fashion and scent experts.

Next, the learning unit 320 receives a plurality of styles and characteristic information already collected in the storage DB 50 as input, and outputs a plurality of first and second flavor series mapped according to the style and characteristic information. It is possible to learn an artificial intelligence-based fashion trend linkage model through machine learning.

Next, the accuracy determination unit 330 applies the feature information detected through the feature detection unit 210 and the user style registered through the style registration unit 230 to the fashion fragrance linkage model, and output values for each scent series to the accuracy level. can be determined by

Next, the scent series registration unit 340 selects at least one scent series having an accuracy of a certain value or more as a user scent series, integrates the user information, and registers the scent series in the storage DB 50 .

Depending on the embodiment, the learning unit 320 may adjust the weight for the fashion scent association model based on the number of purchase history strokes and the number of sample requests of the customized perfume list page generated according to the user scent sequence.

6 is a block diagram of the recommendation unit 500 of FIG. 1 according to an embodiment.

Referring to FIGS. 1 and 6 , the recommendation unit 500 may include a ranking unit 510 , a listing unit 520 and a list providing unit 530 .

First, the ranking unit 510 may automatically determine the list ranking of a plurality of perfume products according to the intensity of fragrance based on the psychological state of the user confirmed through the perfume recommendation service from the user terminal 10 .

Specifically, the ranking unit 510 may detect the user's respiration and heart rate signals through a perfume recommendation service from laser sensors (not shown) provided in the user terminal 10 . At this time, if the user's breathing and heart rate signals are greater than or equal to a preset value, the ranking unit 510 determines the user's psychological state as unstable, and if the user's respiratory and heart rate signals are less than the preset value, the user's psychological state is stable. status can be checked.

Then, when the user's psychological state is confirmed to be in an unstable state, the ranking unit 510 may automatically determine the priority of the plurality of perfume products in the order of high perfume intensity. In addition, when the psychological state of the user is confirmed to be stable, the ranking unit 510 may automatically prioritize the plurality of perfume products in the order of low scent intensity.

Next, the listing unit 520 may list perfume products of a predetermined rank or higher among a plurality of perfume products, and create a customized perfume list page that is payable.

Next, the page providing unit 530 may provide the customized perfume list page generated through the listing unit 520 to the user terminal 10 through a perfume recommendation service.

In this specification, only a few examples of various embodiments performed by the present inventors are described, but the technical spirit of the present invention is not limited or limited thereto, and can be modified and implemented in various ways by those skilled in the art, of course.

10: user terminal
50: storage DB
100: input unit
200: classification unit
300: extraction unit
400: search unit
500: recommendation unit
1000: Fashion-linked personalized perfume recommendation system using AI

Claims (10)

an input unit that receives a coordinating image from a user terminal through a perfume recommendation service;
a classification unit that classifies a user fashion style from a plurality of preset styles based on feature information of an item object detected for each part from the coordinating image;
a selection unit that selects a user scent series corresponding to the user fashion style and the feature information among a plurality of pre-collected scent series using a pre-learned artificial intelligence-based fashion scent linkage model;
a search unit for searching for a plurality of perfume products containing the user fragrance series; and
A recommendation unit for creating a customized perfume list page and recommending it to the user terminal based on product information of each of the plurality of perfume products and whether or not the user fragrance series is the same type;
The input unit draws the coordinating image based on the preferred image for each part selected by the user as sample images are provided through the perfume recommendation service;
The classification unit includes a feature detection unit detecting feature information through object recognition analysis on the item object detected for each part from the coordinating image;
a similarity output unit outputting a similarity to the feature information for each style using a pre-learned artificial intelligence-based style classification model; and
A style registration unit for classifying and registering at least one style having a similarity of a predetermined value or more among the plurality of styles as a user style;
The input unit applies a preset correction filter according to age, gender, and occupation to the coordinating image based on user information received through the perfume recommendation service,
The recommendation unit automatically outputs use review information for each perfume product listed on the customized perfume list page based on a preset facial expression object detected from a facial image captured at regular intervals through a camera provided in the user terminal. let it,
The recommendation unit filters some perfume products from the customized perfume list page based on feedback information extracted from user review information for each perfume product,
The recommender recommends fashion-related customized perfumes using AI, which sets the intensity of each fragrance series corresponding to external and internal activities as the user's preferred intensity based on the noise intensity detected through the microphone provided in the user terminal. system. delete delete According to claim 1,
The similarity output unit learns the style classification model in advance through machine learning that takes a plurality of pre-collected feature information as input and outputs a plurality of pre-collected style categories,
A fashion-linked customized perfume recommendation system using AI, which determines an output value derived by applying the feature information to the style classification model as a similarity to the feature information. According to claim 1,
The selector includes a collection unit that collects a plurality of first scent series mapped by style through fashion and scent experts and a plurality of second scent series mapped by feature information;
An artificial intelligence-based fashion scent linkage model through machine learning that takes a plurality of pre-collected styles and feature information as input and outputs the plurality of first and second fragrance series mapped according to style and feature information. Learning unit for learning;
an accuracy determination unit for determining an output value for each fragrance series outputted as accuracy when the user fashion style and the characteristic information are applied to the fashion fragrance linkage model; and
A perfume recommendation system linked to fashion using AI, including a fragrance series registration unit for selecting and registering at least one fragrance series having the accuracy of a certain value or higher as a user scent series. According to claim 5,
The learning unit adjusts the weight for the fashion fragrance association model based on the number of purchase history and the number of sample requests of the customized perfume list page generated according to the user fragrance series. A customized perfume recommendation system linked to fashion using AI. According to claim 1,
a ranking unit that automatically determines a list ranking of the plurality of perfume products according to scent sequence intensity based on the user's psychological state confirmed from the user terminal through the perfume recommendation service;
a listing unit that lists perfume products having a predetermined rank or higher among the plurality of perfume products and creates a customized perfume list page that is payable; and
A fashion-linked customized perfume recommendation system using AI, including a page providing unit for providing the customized perfume list page to the user terminal through the perfume recommendation service. According to claim 7,
When the user's psychological state is confirmed to be in an unstable state, the ranking unit determines the list ranking of the plurality of perfume products in the order of fragrance sequence intensity,
When the user's psychological state is confirmed to be in a stable state, a fashion-linked customized perfume recommendation system using AI, which determines the list ranking of the plurality of perfume products in the order of low fragrance series intensity. According to claim 1,
The input unit searches for a fashion image for each part corresponding to the text information received from the user terminal through the perfume recommendation service,
A fashion-linked customized perfume recommendation system using AI that draws a coordinating image using the fashion image for each part. According to claim 1,
The recommendation unit adjusts the perfume duration for each perfume product in inverse proportion to the body temperature information received from the user terminal through the perfume recommendation service, and guides the reuse time according to the perfume use time in the form of an alarm, AI. A personalized perfume recommendation system linked to fashion.

Images