Commodity recommendation method and system based on artificial intelligence and big data processing

Abstract

The present invention provides a commodity recommendation method and system based on artificial intelligence and big data processing, and the present invention relates to the technical field of commodity recommendation. The method comprises acquiring a user face image; obtaining a plurality of user face images after lipstick makeup using a generative adversarial network based on the user face image; displaying the plurality of user face images after lipstick makeup on a mobile phone screen; acquiring a target user face image after lipstick makeup selected by the user; extracting expected lipstick makeup information using a makeup extraction model based on the target user face image after lipstick makeup and the user face image; determining a plurality of lipstick cosmetic links based on the expected lipstick makeup information using a recommendation model; calculating the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link; and recommending a plurality of lipstick cosmetic links whose similarity is greater than a similarity threshold to the user. The method can accurately recommend lipstick cosmetics suitable for the user.

Description

Technical Field

[0001] The present invention relates to the technical field of commodity recommendation, and in particular to a commodity recommendation method and system based on artificial intelligence and big data processing. Background Art

[0002] In the modern cosmetics market, consumers are faced with numerous brands and product choices, and finding the lipstick color and style that best suits them is often a challenge. Traditional cosmetics recommendation methods rely on manual color testing and sales staff's advice, which is inefficient and subjective, and difficult to meet consumers' personalized needs. Existing cosmetics recommendation methods on e-commerce platforms also have some limitations. First, these systems often rely on user-entered preference information, such as skin color, lip color, etc., which may be inaccurate or incomplete, resulting in the recommended lipstick color not matching the user's actual needs.

[0003] Therefore, how to accurately recommend lipstick cosmetics suitable for users is a problem that needs to be solved urgently. Summary of the invention

[0004] The main technical problem solved by the present invention is how to accurately recommend lipstick cosmetics suitable for users.

[0005] According to a first aspect, the present invention provides a commodity recommendation method based on artificial intelligence and big data processing, comprising: obtaining a user face image; obtaining multiple user face images after lipstick makeup based on the user face image using a generative adversarial network; displaying the multiple user face images after lipstick makeup on a mobile phone screen; obtaining a target user face image after lipstick makeup selected by the user; extracting expected lipstick makeup information based on the target user face image after lipstick makeup and the user face image using a makeup extraction model; determining multiple lipstick cosmetic links based on the expected lipstick makeup information using a recommendation model; calculating the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link; and recommending multiple lipstick cosmetic links whose similarity is greater than a similarity threshold to the user.

[0006] Furthermore, the calculation of the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetics link includes: obtaining the before and after makeup images of the commenting user in the comment information of the multiple lipstick cosmetics links; extracting the lipstick makeup information of each user in the comment information using the makeup extraction model based on the before and after makeup images of the commenting user in the comment information of each lipstick cosmetics link; calculating multiple first similarities based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information; accumulating the multiple first similarities and then dividing them by the total number of users in the comment information to obtain the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetics link.

[0007] Furthermore, the calculating of multiple first similarities based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information includes: calculating the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information, and calculating the first similarity between the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information through the Hamming distance.

[0008] Furthermore, the makeup extraction model is a convolutional neural network model, the input of the makeup extraction model is the target user's face image after lipstick makeup and the user's face image, and the output of the makeup extraction model is the expected lipstick makeup information.

[0009] Furthermore, the recommendation model is an artificial neural network model, the input of the recommendation model is the lipstick makeup information, and the output of the recommendation model is a plurality of lipstick cosmetic links.

[0010] According to a second aspect, the present invention provides a commodity recommendation system based on artificial intelligence and big data processing, comprising: a first acquisition module, for acquiring a user face image;

[0011] A lipstick makeup generation module, used to obtain multiple user face images with lipstick makeup based on the user face image using a generative adversarial network;

[0012] A display module, used for displaying the plurality of lipstick-made-up user face images on a mobile phone screen;

[0013] The second acquisition module is used to acquire the target user's face image after the user selects the lipstick makeup;

[0014] A lipstick makeup information determination module is used to extract expected lipstick makeup information based on the target user's face image after lipstick makeup and the user's face image using a makeup extraction model;

[0015] A lipstick cosmetics link determination module, configured to determine a plurality of lipstick cosmetics links using a recommendation model based on the expected lipstick makeup information;

[0016] A similarity calculation module, used to calculate the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link;

[0017] The recommendation module is used to recommend multiple lipstick cosmetic links with similarity greater than a similarity threshold to the user.

[0018] Furthermore, the similarity calculation module is further used to: obtain the before and after makeup images of the commenting user in the comment information of the multiple lipstick cosmetics links;

[0019] Based on the before-and-after makeup images of the commenting user in the comment information of each lipstick cosmetic link, the makeup extraction model is used to extract the lipstick makeup information of each user in the comment information;

[0020] Calculating a plurality of first similarities based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information;

[0021] The multiple first similarities are accumulated and then divided by the total number of users in the comment information to obtain the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link.

[0022] Furthermore, the similarity calculation module is also used for:

[0023] The SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information are calculated, and the first similarity between the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information is obtained through Hamming distance calculation.

[0024] Furthermore, the makeup extraction model is a convolutional neural network model, the input of the makeup extraction model is the target user's face image after lipstick makeup and the user's face image, and the output of the makeup extraction model is the expected lipstick makeup information.

[0025] Furthermore, the recommendation model is an artificial neural network model, the input of the recommendation model is the lipstick makeup information, and the output of the recommendation model is a plurality of lipstick cosmetic links.

[0026] The present invention provides a commodity recommendation method and system based on artificial intelligence and big data processing, the method comprising acquiring a user face image; obtaining a plurality of user face images after lipstick makeup using a generative adversarial network based on the user face image; displaying the plurality of user face images after lipstick makeup on a mobile phone screen; acquiring a target user face image after lipstick makeup selected by the user; extracting expected lipstick makeup information using a makeup extraction model based on the target user face image after lipstick makeup and the user face image; determining a plurality of lipstick cosmetic links based on the expected lipstick makeup information using a recommendation model; calculating the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link; and recommending a plurality of lipstick cosmetic links whose similarity is greater than a similarity threshold to the user. The method can accurately recommend lipstick cosmetics suitable for the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flowchart of a commodity recommendation method based on artificial intelligence and big data processing provided by an embodiment of the present invention;

[0028] Figure 2 A schematic diagram of a process for calculating the similarity between expected lipstick makeup information and lipstick makeup information in each lipstick cosmetic link provided by an embodiment of the present invention;

[0029] Figure 3 A schematic diagram of a product recommendation system based on artificial intelligence and big data processing provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In an embodiment of the present invention, there is provided Figure 1 A commodity recommendation method based on artificial intelligence and big data processing is shown, and the commodity recommendation method based on artificial intelligence and big data processing includes steps S1 to S8:

[0031] Step S1, obtaining a user's face image.

[0032] In some embodiments, a user's face photo or real-time video frame can be obtained from the user's mobile phone album or camera as the user's face image.

[0033] Step S2: obtaining a plurality of user face images with lipstick makeup using a generative adversarial network based on the user face image.

[0034] The input of the generative adversarial network is the user face image, and the output of the generative adversarial network is a plurality of user face images after lipstick makeup. The generative adversarial network (GAN) consists of a generator and a discriminator. These two parts confront and learn from each other and jointly promote the training of the model. In the scenario of obtaining a user face image after lipstick makeup based on the user face image, the generator receives the user face image as input and generates a forged sample after lipstick makeup. The discriminator receives the forged sample after lipstick makeup and the real sample (such as the real photo of other users after lipstick makeup) and tries to distinguish them. The generator optimizes its generation ability according to the feedback of the discriminator, so that the forged sample after lipstick makeup is closer to the real sample. The discriminator is updated according to the new sample generated by the generator to improve the discrimination ability of the forged sample after lipstick makeup. The above steps are iterated repeatedly until the forged sample after lipstick makeup generated by the generator is realistic enough and the discriminator cannot distinguish between the real sample and the forged sample. In this way, the generative adversarial network can obtain multiple user face images with lipstick makeup based on the user face image. The generator learns the relationship between the user face image and the lipstick makeup, and can generate forged samples with different lipstick colors and styles. The discriminator helps the generator to continuously improve its generation ability, making the generated forged samples with lipstick makeup more realistic. In this way, the generative adversarial network can be used to generate multiple user face images with different lipstick makeup effects for users to choose and refer to.

[0035] The multiple lipstick makeup-applied user face images are multiple lipstick makeup-applied images obtained by synthesizing the user face images with different lipstick makeups using a generative adversarial network.

[0036] By showing multiple images of users’ faces with lipstick makeup applied, users can intuitively understand the effects of different lipstick colors and styles on their appearance. This helps users choose the lipstick that suits them more accurately.

[0037] Step S3, displaying the plurality of lipstick-made-up user face images on the mobile phone screen.

[0038] The plurality of lipstick-made-up user face images can be displayed on a mobile phone screen for user selection.

[0039] In some embodiments, the displaying of the plurality of user face images after lipstick makeup on the mobile phone screen includes: processing the plurality of user face images after lipstick makeup by a sorting model to obtain display serial numbers of the plurality of user face images after lipstick makeup, and displaying the plurality of user face images after lipstick makeup on the mobile phone screen based on the display serial numbers of the plurality of user face images after lipstick makeup, wherein the sorting model is a convolutional neural network model, the input of the sorting model is the plurality of user face images after lipstick makeup, and the output of the sorting model is the serial numbers of the plurality of user face images after lipstick makeup. The convolutional neural network model is an implementation of artificial intelligence. A convolutional neural network (CNN) may be a multi-layer neural network (e.g., including at least two layers). The at least two layers may include at least one of a convolutional layer (CONV), a rectified linear unit (ReLU) layer, a pooling layer (POOL), or a fully connected layer (FC). A convolutional neural network can extract useful features from an image and gradually understand and learn the contextual information of the image. The sorting model can sort the multiple user face images after lipstick makeup according to the makeup effects of the multiple user face images after lipstick makeup, and the smaller the sequence number, the better the makeup effect. For example, the sequence number is 1, which means the best makeup effect, and the picture will also be displayed in the first position on the mobile phone screen.

[0040] Step S4, obtaining the target user's face image after applying the lipstick makeup selected by the user.

[0041] The user can select one of the multiple user face images after lipstick makeup as the target user face image after lipstick makeup selected by the user. The target user face image after lipstick makeup selected by the user represents the image of the lipstick makeup effect that the user is most satisfied with.

[0042] Step S5, extracting expected lipstick makeup information using a makeup extraction model based on the target user's face image after lipstick makeup and the user's face image.

[0043] The makeup extraction model is a convolutional neural network model, the input of the makeup extraction model is the target user's face image after lipstick makeup and the user's face image, and the output of the makeup extraction model is the expected lipstick makeup information.

[0044] The expected lipstick makeup information is the lipstick makeup information extracted by the makeup extraction model based on the target user face image and the user face image uploaded by the user. The expected lipstick makeup information includes lipstick color, lipstick texture, and lipstick coverage. Lipstick textures include matte, velvet, pearlescent, etc. Lipstick coverage indicates the degree of coverage of the lips. The higher the lipstick coverage, the higher the degree to which the original color of the lips is completely covered. The convolutional neural network can accurately extract the expected lipstick makeup information by learning the local features in the image.

[0045] In some embodiments, the makeup extraction model includes an image segmentation layer, an image matching layer, a lipstick image extraction layer, and an expected lipstick makeup information determination layer. The image segmentation layer, the image matching layer, the lipstick image extraction layer, and the expected lipstick makeup information determination layer all include a convolutional neural network. The input of the image segmentation layer is the target user face image after lipstick makeup and the user face image, the output of the image segmentation layer is the user lip image after lipstick makeup and the user lip image without lipstick makeup, the input of the image matching layer is the user lip image after lipstick makeup and the user lip image without lipstick makeup, the output of the image matching layer is the matching matrix of the user lip image after lipstick makeup and the user lip image without lipstick makeup, the input of the lipstick image extraction layer is the matching matrix of the user lip image after lipstick makeup and the user lip image without lipstick makeup, the user lip image after lipstick makeup and the user lip image without lipstick makeup, the output of the lipstick image extraction layer is the lipstick image, the input of the expected lipstick makeup information determination layer is the lipstick image, the user lip image after lipstick makeup and the user lip image without lipstick makeup, the matching matrix of the user lip image after lipstick makeup and the user lip image without lipstick makeup, and the output of the expected lipstick makeup information determination layer is the expected lipstick makeup information. Each element in the matching matrix of the user's lip image after lipstick makeup and the user's lip image without lipstick makeup represents the degree of matching between a position in the user's lip image after lipstick makeup and a corresponding position in the user's lip image without lipstick makeup. Such a matrix can be used to represent the pixel-level matching relationship between the two images. The image segmentation layer performs image segmentation through a convolutional neural network to separate the user's lip image after lipstick makeup and the user's lip image without lipstick makeup from the overall image. The image matching layer calculates the matching matrix between the two images, in which each element represents the degree of matching between a position in the user's lip image after lipstick makeup and a corresponding position in the user's lip image without lipstick makeup. Through the matching matrix, the pixel-level matching relationship between the two images can be understood. The lipstick image extraction layer extracts the lipstick image from the input through a convolutional neural network. The output of this layer is a lipstick image, that is, the lipstick part is extracted from the user's lip image. The expected lipstick makeup information determination layer determines the specific information of the expected lipstick makeup based on information such as the lipstick image and the matching matrix.

[0046] Step S6: determining a plurality of lipstick cosmetic links using a recommendation model based on the lipstick makeup information.

[0047] The recommendation model is an artificial neural network model, the input of the recommendation model is the lipstick makeup information, and the output of the recommendation model is a plurality of lipstick cosmetic links. The basic structure of the artificial neural network model includes an input layer, a hidden layer, and an output layer. The input layer receives external input data, the hidden layer is responsible for processing and processing the input data, and finally passes the result to the output layer, which generates the final prediction or classification result. The artificial neural network model can handle nonlinear relationships, and the lipstick makeup information usually contains complex relationships between multiple features. For example, features such as lipstick color, texture, and coverage may affect each other, and the artificial neural network model can capture these complex relationships through the combination of multiple layers of neurons and the nonlinear transformation of the activation function to screen out multiple lipstick cosmetic links.

[0048] Step S7, calculating the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link.

[0049] In some embodiments, the Figure 2 To calculate the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetics link, Figure 2 A flow chart of calculating the similarity between expected lipstick makeup information and lipstick makeup information in each lipstick cosmetics link provided by an embodiment of the present invention, wherein the calculating the similarity between expected lipstick makeup information and lipstick makeup information in each lipstick cosmetics link comprises steps S21 to S24:

[0050] Step S21, obtaining the before and after makeup images of the commenting user from the comment information of the plurality of lipstick cosmetics links.

[0051] The before-and-after makeup images of the commenting user in the comment information of the multiple lipstick cosmetics links refer to the facial photos of the user taken before and after using the lipstick cosmetics. For example, the photos taken by the user before and after using the lipstick. As an example, the comment information of the lipstick cosmetics link contains the evaluation of the before-and-after makeup images of 10 users, and the before-and-after makeup images of these users can be obtained for subsequent lipstick makeup information extraction and similarity calculation.

[0052] Step S22, based on the before and after makeup images of the commenting user in the comment information of each lipstick cosmetic link, use the makeup extraction model to extract the lipstick makeup information of each user in the comment information.

[0053] For the makeup extraction model, please refer to the relevant description of step S5, which will not be repeated here.

[0054] Step S23, calculating a plurality of first similarities based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information.

[0055] In some embodiments, the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information can be calculated, and multiple first similarities between the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information can be obtained through Hamming distance calculation.

[0056] Step S24, accumulating the multiple first similarities and dividing the sum by the total number of users in the comment information to obtain the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link.

[0057] By accumulating multiple first similarities and dividing them by the total number of users in the comment information, a more accurate similarity calculation result can be obtained, thereby improving the accuracy of lipstick recommendation. After accumulating multiple first similarities and dividing them by the total number of users in the comment information, the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetics link is obtained. This similarity value can be used to measure the degree of match between each lipstick cosmetics link and user needs.

[0058] Step S8: recommend multiple lipstick cosmetics links with similarities greater than a similarity threshold to the user.

[0059] By calculating the similarity and filtering out lipstick cosmetics links with similarity greater than the threshold, it can ensure that the recommended lipstick is more closely matched with the user's expected lipstick makeup information, and can provide more accurate lipstick recommendation results. Lipstick cosmetics links with similarity greater than the threshold represent lipsticks that are closer to the user's needs. Narrowing the lipstick selection range to lipstick cosmetics links with similarity greater than the threshold can help users quickly find products that meet their lipstick needs, saving time and energy. The similarity threshold can be manually entered in advance.

[0060] Based on the same inventive concept, Figure 3 A schematic diagram of a commodity recommendation system based on artificial intelligence and big data processing provided by an embodiment of the present invention, the commodity recommendation system based on artificial intelligence and big data processing includes:

[0061] A first acquisition module 31, used to acquire a user's face image;

[0062] A lipstick makeup generation module 32 is used to obtain multiple user face images with lipstick makeup based on the user face image using a generative adversarial network;

[0063] A display module 33, used for displaying the plurality of lipstick-made-up user face images on a mobile phone screen;

[0064] The second acquisition module 34 is used to acquire the target user's face image after the user has applied the lipstick selected by the user;

[0065] A lipstick makeup information determination module 35 is used to extract expected lipstick makeup information using a makeup extraction model based on the target user's face image after lipstick makeup and the user's face image;

[0066] A lipstick cosmetics link determination module 36 is used to determine a plurality of lipstick cosmetics links using a recommendation model based on the expected lipstick makeup information;

[0067] A similarity calculation module 37, used to calculate the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link;

[0068] The recommendation module 38 is used to recommend a plurality of lipstick cosmetic links having a similarity greater than a similarity threshold to the user.

Claims

1. A commodity recommendation method based on artificial intelligence and big data processing, characterized in that: include: Get the user's face image; Based on the user face image, a generative adversarial network is used to obtain multiple user face images with lipstick makeup; Displaying the plurality of lipstick-made-up user face images on a mobile phone screen; Obtain the target user's face image after the user selects the lipstick makeup; Extracting expected lipstick makeup information using a makeup extraction model based on the target user face image after lipstick makeup and the user face image, wherein the makeup extraction model is a convolutional neural network model, and the input of the makeup extraction model is the target user face image after lipstick makeup and the user face image, and the output of the makeup extraction model is the expected lipstick makeup information, and the expected lipstick makeup information includes lipstick color, lipstick texture, and lipstick coverage; Determine a plurality of lipstick cosmetic links using a recommendation model based on the expected lipstick makeup information; Calculating the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetics link, wherein calculating the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetics link comprises: Obtaining the before-and-after makeup images of the commenting user from the comment information of the multiple lipstick cosmetics links; Based on the before-and-after makeup images of the commenting user in the comment information of each lipstick cosmetic link, the makeup extraction model is used to extract the lipstick makeup information of each user in the comment information; The first similarities are calculated based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information, wherein the first similarities are calculated based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information, including: Calculating the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information, and obtaining a first similarity between the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information through Hamming distance calculation; The multiple first similarities are accumulated and then divided by the total number of users in the comment information to obtain the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link; Recommend multiple lipstick cosmetics links whose similarity is greater than a similarity threshold to the user.

2. The commodity recommendation method based on artificial intelligence and big data processing according to claim 1, characterized in that: The recommendation model is an artificial neural network model, the input of the recommendation model is the lipstick makeup information, and the output of the recommendation model is a plurality of lipstick cosmetic links.

3. A product recommendation system based on artificial intelligence and big data processing, characterized in that: include: The first acquisition module is used to acquire a user's face image; A lipstick makeup generation module, used to obtain multiple user face images with lipstick makeup based on the user face image using a generative adversarial network; A display module, used for displaying the plurality of lipstick-made-up user face images on a mobile phone screen; The second acquisition module is used to acquire the target user's face image after the user selects the lipstick makeup; a lipstick makeup information determination module, configured to extract expected lipstick makeup information using a makeup extraction model based on the target user face image after lipstick makeup and the user face image, wherein the makeup extraction model is a convolutional neural network model, the input of the makeup extraction model is the target user face image after lipstick makeup and the user face image, and the output of the makeup extraction model is the expected lipstick makeup information, wherein the expected lipstick makeup information includes lipstick color, lipstick texture, and lipstick coverage; A lipstick cosmetics link determination module, configured to determine a plurality of lipstick cosmetics links using a recommendation model based on the expected lipstick makeup information; A similarity calculation module is used to calculate the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link, and the similarity calculation module is also used to: Obtaining the before-and-after makeup images of the commenting user from the comment information of the multiple lipstick cosmetics links; Based on the before-and-after makeup images of the commenting user in the comment information of each lipstick cosmetic link, the makeup extraction model is used to extract the lipstick makeup information of each user in the comment information; Based on the expected lipstick makeup information and the lipstick makeup information of each user in the comment information, a plurality of first similarities are calculated, and the similarity calculation module is further used to: Calculating the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information, and obtaining a first similarity between the SimHash value of the expected lipstick makeup information and the SimHash value of the lipstick makeup information of each user in the comment information through Hamming distance calculation; The multiple first similarities are accumulated and then divided by the total number of users in the comment information to obtain the similarity between the expected lipstick makeup information and the lipstick makeup information in each lipstick cosmetic link; The recommendation module is used to recommend multiple lipstick cosmetic links with similarity greater than a similarity threshold to the user.

4. The commodity recommendation system based on artificial intelligence and big data processing as claimed in claim 3, characterized in that: The recommendation model is an artificial neural network model, the input of the recommendation model is the lipstick makeup information, and the output of the recommendation model is a plurality of lipstick cosmetic links.

Images