Makeup assisting method and system based on artificial intelligence large model

Abstract

The invention provides a makeup assistance method and system based on an artificial intelligence large model, and belongs to the technical field of intelligent makeup assistance, and the method comprises the steps: determining a user portrait based on multi-source user information, and determining a makeup assistance mode based on the user portrait. And determining an initial recommended makeup based on the user portrait, the scene type and the first face information, and generating first makeup guidance information based on the makeup auxiliary mode and the initial recommended makeup. The first face information is the face information of the user before makeup. And adjusting the initial recommended makeup based on the beauty makeup large model, the second face information and the first user feedback information to obtain a target recommended makeup. And adjusting the first beauty makeup guidance information based on the target recommended makeup to obtain second beauty makeup guidance information. The second face information is the face information of the user after makeup. The accuracy of beauty makeup recommendation and auxiliary guidance can be improved, and user requirements are met.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of intelligent beauty makeup assistance, and more specifically, relates to a beauty makeup assistance method and system based on an artificial intelligence large model. Background Art

[0002] With the improvement of people's living standards, the scale of the beauty makeup market has been continuously expanding. Nowadays, beauty makeup is no longer the patent of women only. More and more men have begun to pay attention to and participate in beauty makeup consumption, which has further promoted the diversified development of the beauty makeup market. Consumers' demands for beauty makeup products and services are no longer limited to traditional basic functions, but are increasingly moving towards diversification and personalization. Due to the large differences in the mastery of beauty makeup knowledge and skills among the consumer group, there is a greater need for a comprehensive and intelligent beauty makeup assistance method.

[0003] However, the existing beauty makeup assistance methods and beauty makeup recommendation methods are difficult to quickly adapt to the dynamic changes of beauty makeup trends and users' skin types and preferences, and it is difficult to provide accurate personalized beauty makeup guidance adapted to users during long-term application. At the same time, there is also a lack of an effective mechanism to automatically learn new makeup looks and user feedback to update the model, and it may not be able to meet the increasingly diverse needs of users over time. Summary of the Invention

[0004] The purpose of the present disclosure is to provide a beauty makeup assistance method and system based on an artificial intelligence large model to improve the accuracy of beauty makeup recommendation and assistance guidance and meet the needs of users.

[0005] In the first aspect of the embodiments of the present disclosure, a beauty makeup assistance method based on an artificial intelligence large model is provided, including: Determining a user portrait based on multi-source user information, and determining a beauty makeup assistance mode based on the user portrait.

[0006] Determining an initial recommended makeup look based on the user portrait, scene type, and first face information, and generating first beauty makeup guidance information based on the beauty makeup assistance mode and the initial recommended makeup look. The first face information is the face information of the user before applying makeup.

[0007] Adjusting the initial recommended makeup look based on the beauty makeup large model, second face information, and first user feedback information to obtain a target recommended makeup look. Adjusting the first beauty makeup guidance information based on the target recommended makeup look to obtain second beauty makeup guidance information. The second face information is the face information of the user after applying makeup.

[0008] In the second aspect of the embodiments of the present disclosure, a beauty makeup assistance system based on an artificial intelligence large model is provided, including: A user data analysis module, configured to determine a user portrait based on multi-source user information, and determine a beauty makeup assistance mode based on the user portrait.

[0009] The makeup recommendation module is used to determine an initial recommended makeup based on the user profile, scene type, and first facial information, and generate first makeup guidance information based on the makeup assistance mode and the initial recommended makeup. The first facial information is the facial information of the user before applying makeup.

[0010] The feedback optimization module is used to adjust the initial recommended makeup based on the makeup large model, second facial information, and first user feedback information to obtain a target recommended makeup. Adjust the first makeup guidance information based on the target recommended makeup to obtain second makeup guidance information. The second facial information is the facial information of the user after applying makeup.

[0011] In the third aspect of the embodiments of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the above-mentioned makeup assistance method based on the artificial intelligence large model are implemented.

[0012] In the fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the above-mentioned makeup assistance method based on the artificial intelligence large model are implemented.

[0013] The beneficial effects of the makeup assistance method and system based on the artificial intelligence large model provided by the embodiments of the present disclosure are as follows: In this embodiment, by integrating multi-source user information to construct an accurate user profile, the unique needs and preferences of users can be deeply understood. Matching the makeup assistance mode according to the user profile, providing exclusive services for users with different experience levels, whether it is detailed guidance for novices or efficient suggestions for experienced users, can meet their personalized needs.

[0014] In this embodiment, the initial recommended makeup is screened by combining the user profile, scene type, and facial information, fully considering the user characteristics and usage scenarios, so that the recommended makeup is more in line with the actual needs. Using advanced algorithms to ensure the accuracy and rationality of makeup recommendations, and improving the satisfaction of users with the recommended makeup.

[0015] In this embodiment, with the help of the makeup large model, the post-makeup image, and user feedback, the initial recommended makeup and guidance information are intelligently adjusted. It can not only optimize the makeup according to the actual effect, but also update the guidance information in real time, help users continuously improve their makeup, provide dynamic and considerate makeup assistance services, and enhance the user's makeup experience. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a schematic flowchart of a beauty makeup assistance method based on an artificial intelligence large model provided by an embodiment of the present disclosure; Figure 2 It is a functional structure diagram of a beauty makeup assistance method based on an artificial intelligence large model provided by an embodiment of the present disclosure; Figure 3 It is a schematic structural diagram of a beauty makeup large model provided by an embodiment of the present disclosure; Figure 4 It is a schematic flowchart of another beauty makeup assistance method based on an artificial intelligence large model provided by an embodiment of the present disclosure; Figure 5 It is a structural block diagram of a beauty makeup assistance system based on an artificial intelligence large model provided by an embodiment of the present disclosure; Figure 6 It is a schematic block diagram of an electronic device provided by an embodiment of the present disclosure. Detailed implementation manners

[0018] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are presented to thoroughly understand the embodiments of the present disclosure. However, those skilled in the art should clearly understand that the present disclosure can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present disclosure.

[0019] To make the purpose, technical solutions, and advantages of the present disclosure clearer, the following will be described through specific embodiments in conjunction with the drawings.

[0020] Please refer to Figures 1 to 4 , Figure 1 It is a schematic flowchart of a beauty makeup assistance method based on an artificial intelligence large model provided by an embodiment of the present disclosure, and this method may include S101 to S103.

[0021] S101: Determine a user profile based on multi-source user information, and determine a beauty makeup assistance mode based on the user profile.

[0022] In this embodiment, the multi-source user information includes user image information, user basic information, user application behavior information, and user voice information.

[0023] Determining a user portrait based on multi-source user information, including: Determining a first user portrait based on the user's basic information and voice information.

[0024] Extracting the user's facial features and makeup features based on the user's image information.

[0025] Determining the preference features of the makeup style based on the user's application behavior information.

[0026] Embedding the user's facial features, makeup features and makeup style preference features into the first user portrait to obtain the user portrait.

[0027] Figure 2 This is a functional structure diagram of a makeup assistance method based on an artificial intelligence large model provided by an embodiment of the present disclosure. Refer to Figure 2 , in this embodiment, a makeup assistance method based on an artificial intelligence large model can be applied to a makeup mirror. A detachable camera, microphone and speaker are installed on the makeup mirror. When the user uses the makeup mirror to apply makeup, this method can provide makeup assistance services for the user. The makeup assistance services can include personalized makeup recommendations, real-time makeup try-on (i.e., virtual makeup try-on) and voice interaction.

[0028] Figure 3 This is a schematic structural diagram of a makeup large model provided by an embodiment of the present disclosure. Refer to Figure 3 , the artificial intelligence large model is a makeup large model. The makeup large model can include a facial data large model, a historical data large model, an emotion data large model, a speech recognition large model, a speech synthesis large model and a speech wake-up large model. Among them, the facial data large model can accurately analyze the user's features based on the user's facial information collected by the camera, provide personalized makeup recommendations that fit the facial characteristics of the user, and ensure that the virtual makeup fits perfectly with the face during virtual makeup try-on, improving the realism of the makeup try-on.

[0029] Exemplarily, the historical data large model stores the user's historical application behavior information related to makeup and the open-source makeup data. By analyzing these historical data, the user's makeup preferences and habits are obtained, and makeup and products that meet their preferences are pushed to the user to achieve personalized services.

[0030] The emotion data large model can analyze the user's facial expression based on the user's face information collected by the camera, and judge the user's emotional reaction to the makeup. If the user shows dissatisfaction, the recommendation is adjusted in time; if the user is satisfied, the relevant recommendation is strengthened to make the makeup assistance more in line with the user's psychological needs.

[0031] The speech recognition large model is used to convert the user's voice command into text and analyze the user's intention.

[0032] The large language model for speech synthesis can convert beauty makeup guidance and recommendation information into speech, which is played through a speaker to give users clear and easy-to-understand voice prompts, enhancing the user experience.

[0033] The large language model for voice wake-up is used to constantly monitor sounds, identify preset wake-up words, and quickly activate the voice interaction function without manual operation, improving the convenience of use.

[0034] In this embodiment, the user portrait is a virtual image or model constructed by collecting and analyzing multi-dimensional data related to the user (i.e., multi-source user information), which can comprehensively and accurately describe the user's characteristics and behavior patterns. The multi-source user information can be sourced from the information uploaded by the user and the information transmitted by networked devices. The networked devices can be the user's smartphone, tablet, etc. The user's basic information can include basic attribute information such as the user's age, gender, occupation, skin type, skin color, region, etc. The user's voice information can include the audio clips recorded by the user. The user's voiceprint features can be extracted based on the user's voice information, and the user's needs and preferences for beauty makeup can also be analyzed based on the semantics of the voice.

[0035] The first user portrait is a portrait initially constructed based on the user's basic information and voice information, integrating the key points of the basic information and voice information. The user's image information includes the user's facial images, such as self-taken photos, video screenshots, etc., which can be used to analyze the user's facial contour, facial features, skin condition, and existing makeup effect. The user's facial features include the physiological features of the human face, such as face shape, facial feature ratio, skin texture, facial blemishes, etc. The makeup features include the user's makeup characteristics extracted from the image, such as eye makeup style, lip makeup color and texture, blush position and tone, makeup intensity, etc.

[0036] The user's application behavior information includes the interaction records of the user in the beauty makeup application, such as beauty makeup browsing records, purchase records, search keywords, etc. The beauty makeup style preference feature is the user's preference tendency for different beauty makeup styles obtained through the analysis of application behavior. The beauty makeup style preference feature can include styles such as natural and retro, and each style corresponds to a quantified numerical value of different preference degrees.

[0037] In this embodiment, the analysis of multi-source users and the construction of user portraits can be realized by the large language model for facial data, the large language model for historical data, and the large language model for emotion data.

[0038] Exemplarily, a facial data large model can extract facial feature vectors from the user's facial images collected by a camera through pre-trained models such as VGGNet and ResNet. Based on the facial feature vectors, on the one hand, by comparing with the preset makeup-facial feature matching rules, personalized makeup that suits the facial characteristics can be recommended for the user; on the other hand, during virtual makeup try-on, using 3D face reconstruction technology, the virtual makeup can be accurately mapped onto the user's facial 3D model to achieve a perfect fit between the virtual makeup and the user's facial 3D model.

[0039] The emotion data large model can include a facial expression analysis model based on a convolutional neural network. By training on a large amount of face image data with expression annotations, it learns the characteristic patterns of different expressions, thereby identifying the user's facial expressions (such as frowning indicating dissatisfaction, smiling indicating satisfaction, etc.). The emotion data large model can also apply speech emotion analysis technology. By extracting the acoustic features and semantic features of speech, and using support vector machine and recurrent neural network algorithms for emotion classification, it can judge the user's emotional reaction to the makeup. Combining the results of facial expression and speech emotion analysis, the beauty makeup recommendation strategy can be adjusted in real time to make the beauty makeup assistance more in line with the user's psychological needs.

[0040] The historical data large model can use a database management system to store the user's historical information related to beauty makeup and open-source beauty makeup data. Through the Apriori association rule mining algorithm and clustering algorithm, it analyzes the association relationships between different beauty makeup elements in the user behavior data and mines the user behavior patterns and preferences.

[0041] Exemplarily, the beauty makeup large model can determine the age, gender, occupation, skin type, and skin color based on the user's basic information, and construct a user basic profile based on the age, gender, occupation, skin type, and skin color. Based on speech recognition technology, the user's voice information is converted into text information, and based on natural language processing algorithms, the user's needs, preferences, and feedback on beauty makeup are determined. The user's voiceprint features are extracted based on the user's voice information. A first user portrait is constructed based on the user basic profile, the user's voiceprint features, the user's needs, preferences, and feedback on beauty makeup.

[0042] The user's image information is input into a convolutional neural network to extract the user's facial features and makeup features. Based on the Apriori association rule mining algorithm, the user's application behavior information is analyzed to obtain the user's beauty makeup style preference features. Based on the data fusion algorithm, the user's facial features, makeup features, and beauty makeup style preference features are fused to obtain fusion features. Based on the weighted average method, the fusion features are embedded into the first user portrait to obtain the user portrait.

[0043] Exemplarily, the beauty large model collects the user's image information, basic information, application behavior information, and voice information through beauty applications or platforms. Perform preprocessing such as noise reduction and normalization on the image information; perform operations such as cleaning and word segmentation on the text information; perform structured processing on the application behavior data for subsequent analysis. Use pre-trained models such as VGGNet and ResNet to extract the user's facial features and makeup feature vectors based on the user's image information. Analyze the association rules between different beauty elements and styles in the application behavior data based on the Apriori algorithm to determine the beauty style preference feature vector of the user.

[0044] Adopt a weighted average fusion algorithm, assign weights according to the importance of each feature, fuse the first user portrait feature vector with the facial features, makeup features, and beauty style preference feature vectors to obtain a complete user portrait vector and store it in the database for subsequent determination of the beauty makeup assistance mode based on the user portrait.

[0045] S102: Determine the initial recommended makeup based on the user portrait, scene type, and the first face information, and generate the first beauty makeup guidance information based on the beauty makeup assistance mode and the initial recommended makeup. The first face information is the face information of the user before applying makeup.

[0046] In this embodiment, the scene type is the specific situation in which the user uses makeup. For example, scenes such as daily commuting, dinner parties, outdoor sports, etc., and the weather data in the current scene. The user's geographical location and weather data can be obtained through the weather APP in the user's Internet-connected device.

[0047] The first face information is the facial information of the user before applying makeup, which is used to analyze the user's facial features. The initial recommended makeup is a makeup plan initially recommended to the user based on the first face information. The initial recommended makeup may include information such as global makeup, local makeup, and beauty tools.

[0048] The beauty makeup assistance mode is a guidance method set according to the user's beauty makeup experience. The beauty makeup assistance mode may include a learning mode for beginners and a normal mode for experienced users. The first beauty makeup guidance information is an operation guide to help the user complete the initial recommended makeup. The first beauty makeup guidance information may include text-based makeup steps, beauty product usage instructions, precautions, and image-based makeup step demonstration diagrams and final makeup effect diagrams.

[0049] Exemplarily, screen out the initial recommended makeup from the makeup database based on the user portrait, scene type, and the first face information. For example, considering the workplace scene, for users with oily skin, it is possible to preferentially recommend oil-control and natural base makeup, light eye makeup, etc.

[0050] For novice beauty users, determine the beauty makeup assistance mode as the learning mode, and generate the first beauty makeup guidance information based on the learning mode and the initial recommended makeup look. For example, in the learning mode, the makeup steps are refined, accompanied by detailed written explanations and a large number of example pictures; in the normal mode, the key steps and crucial techniques are highlighted, and a schematic diagram of the beauty makeup steps is given.

[0051] Exemplarily, use a convolutional neural network to extract features from the first facial information to obtain the user's facial features. Based on the user profile and the scene type, apply the collaborative filtering algorithm, refer to the preferences of similar users, and screen the initial recommended makeup look from the makeup database.

[0052] For the learning mode, use natural language generation technology to generate detailed text guidance, and combine image synthesis technology to generate step demonstration pictures. For the normal mode, use the text summarization technology of natural language processing to refine the key steps, and select representative images from the image library to provide accurate and efficient beauty makeup guidance for users.

[0053] S103: Adjust the initial recommended makeup look based on the beauty makeup large model, the second facial information, and the first user feedback information to obtain the target recommended makeup look. Adjust the first beauty makeup guidance information based on the target recommended makeup look to obtain the second beauty makeup guidance information. The second facial information is the facial information of the user after applying makeup.

[0054] In this embodiment, the second facial information is the facial image information of the user after starting to apply makeup according to the initial recommended makeup look. The user's facial image during the makeup application process can be collected in real time by a camera installed on the beauty makeup mirror, and this user's facial image is used as the second facial information. Evaluate the actual makeup effect of the user's makeup application based on the second facial information. Perform user facial expression analysis on the second facial information to obtain the user's emotional feedback information.

[0055] The first user feedback information is the evaluation and improvement suggestions of the user on the initial recommended makeup look after applying makeup. The first user feedback information can include voice-like information input by the user. The target recommended makeup look is a makeup look scheme that is more in line with the user's expectations after the beauty makeup large model adjusts the initial recommended makeup look. For example, adjust the base makeup product and its color number, eye makeup style, lip makeup color, etc. of the initial recommended makeup look to obtain the target recommended makeup look.

[0056] The first beauty makeup guidance information is the operation guide initially provided for the user to complete the initial recommended makeup look, including text and image information. The second beauty makeup guidance information is obtained by adjusting the first beauty makeup guidance information according to the target recommended makeup look, and is the information used to guide the user to achieve the target recommended makeup look.

[0057] Figure 4 Schematic diagram of the process of another beauty makeup assistance method based on an artificial intelligence large model provided in an embodiment of the present disclosure. Refer toFigure 4 In this embodiment, the camera and microphone on the beauty mirror first collect the user's facial data and the information about the occasion the user is about to attend. Additionally, the user's address location and weather data are obtained by connecting to the user's smartphone. Based on the voice wake-up module and voice recognition module in the beauty large model, the information about the occasion the user is about to attend (voice information) is analyzed and processed. The processed occasion information data, the user's facial data, and the user's address location and weather data are input into the facial data large model in the beauty large model. Then, combined with the historical data large model and the emotion data large model, the input data is comprehensively analyzed and processed to recommend a suitable makeup plan for the user. The feedback information of the user is collected to analyze the user's satisfaction with the recommended makeup, and the feedback information and satisfaction of the user are stored in the historical data large model to optimize and update the beauty large model.

[0058] Exemplarily, the MaskR-CNN image recognition algorithm based on deep learning is used to detect the makeup parts, analyze the second facial information, and extract the makeup-related features. Natural language processing technologies such as the BERT model are used to perform sentiment analysis and intent extraction on the first user feedback information.

[0059] The extracted makeup-related features and intents are input into the beauty large model, and the beauty large model adjusts the parameters of the initial recommended makeup based on the reinforcement learning algorithm to generate the target recommended makeup.

[0060] Natural language generation technologies such as GPT-based are used to update the text-based beauty guidance information according to the target recommended makeup, and the image synthesis algorithm based on the generative adversarial network GAN is used to adjust the image-based guidance information, thereby obtaining the second beauty guidance information.

[0061] In this embodiment, adjusting the first beauty guidance information based on the target recommended makeup to obtain the second beauty guidance information includes: Determining multiple target makeup element information based on the target recommended makeup.

[0062] Determining multiple initial makeup element information based on the initial recommended makeup.

[0063] Determining the updated makeup element information based on the multiple target makeup element information and the initial makeup element information.

[0064] Screening out the guidance information corresponding to the updated makeup element information from the first beauty guidance information based on the updated makeup element information, and adjusting the guidance information to obtain the second beauty guidance information.

[0065] In this embodiment, the target makeup element information includes the relevant information of each specific component that constitutes the target recommended makeup, such as the shade of the base makeup, the eyeshadow colors and combinations of the eye makeup, the texture of the lip makeup, etc. The initial makeup element information includes the relevant information of each specific component in the initial recommended makeup. The updated makeup element information is determined by comparing the target makeup element information and the initial makeup element information to identify the changed or newly added makeup element information.

[0066] Exemplarily, the initial recommended makeup that user Zhang initially obtained is a natural makeup suitable for daily commuting, and the initial makeup element information is: A liquid foundation in natural shade 01, light pink eyeshadow, light brown eyebrow pencil, and transparent lip gloss. The first beauty makeup guidance information focuses on this makeup and details how to apply the liquid foundation to achieve a light and natural effect, the blending method of the eyeshadow, as well as the usage skills of the eyebrow pencil and the steps of outlining the eyebrow shape.

[0067] However, during the makeup trial process, Zhang feedback that she wanted a more radiant makeup. The beauty makeup large model generated the target recommended makeup based on the feedback, and its target makeup element information became: liquid foundation in natural shade 01, orange-pink eyeshadow, dark brown eyebrow pencil, and rose-colored lipstick.

[0068] By comparison, the updated makeup element information is different colors of eyeshadow and lipstick, and changing the color of the eyebrow pencil. The beauty makeup large model screens out the guidance information corresponding to these updated elements from the first beauty makeup guidance information, such as readjusting the range and order of eyeshadow blending to fit the new colors, and introducing the application skills of the rose-colored lipstick to make it more long-lasting. Finally, the second beauty makeup guidance information is formed to help Zhang complete a makeup that better meets her needs.

[0069] From the above, it can be concluded that in this embodiment, by integrating multi-source user information to construct an accurate user portrait, the unique needs and preferences of users can be deeply understood. Matching the beauty makeup assistance mode according to the user portrait, providing exclusive services for users with different experience levels, whether it is detailed guidance for novices or efficient suggestions for experienced users, can meet their personalized needs.

[0070] In this embodiment, the initial recommended makeup is screened by combining the user portrait, the scene type, and the facial information, fully considering the user characteristics and usage scenarios, making the recommended makeup more in line with the actual needs. Using advanced algorithms to ensure the accuracy and rationality of makeup recommendations, and improving the user's satisfaction with the recommended makeup.

[0071] In this embodiment, with the help of the beauty makeup large model, the post-application image, and user feedback, the initial recommended makeup and guidance information are intelligently adjusted. It can not only optimize the makeup according to the actual effect, but also update the guidance information in real time, help users continuously improve their makeup, provide dynamic and considerate beauty makeup assistance services, and enhance the user's beauty makeup experience.

[0072] In an embodiment of the present disclosure, the beauty makeup assistance mode includes a learning mode and a normal mode.

[0073] Determining the beauty makeup assistance mode based on the user profile includes: Determining the user's beauty makeup experience level based on the user profile.

[0074] If the user's beauty makeup experience level is a beginner in beauty makeup, the beauty makeup assistance mode is the learning mode.

[0075] If the user's beauty makeup experience level is an advanced beauty makeup user, the beauty makeup assistance mode is the normal mode.

[0076] In this embodiment, the first beauty makeup guidance information includes text-based makeup application guidance information and image-based makeup application guidance information.

[0077] Generating the first beauty makeup guidance information based on the beauty makeup assistance mode and the initial recommended makeup look includes: Determining multiple beauty makeup steps based on the initial recommended makeup look.

[0078] If the beauty makeup assistance mode is the learning mode: Generating multiple beauty makeup breakdown steps based on each beauty makeup step. Generating text-based makeup application guidance information and image-based makeup application guidance information based on each beauty makeup breakdown step.

[0079] If the beauty makeup assistance mode is the normal mode: Generating image-based makeup application guidance information based on each beauty makeup step.

[0080] In this embodiment, the learning mode is a guidance mode for helping novice beauty makeup users gradually master makeup skills. The learning mode can provide detailed beauty makeup breakdown steps, comprehensive text-based makeup application guidance information, and rich image-based makeup application guidance information. The normal mode is a concise guidance mode designed for users with a certain level of beauty makeup experience. The normal mode can provide image-based makeup application guidance information based on beauty makeup steps. The user's beauty makeup experience level is a classification of the user's beauty makeup knowledge and skill level. A beginner in beauty makeup represents that the user has just started to contact beauty makeup and has less skill and knowledge reserve. An advanced beauty makeup user represents that the user has a certain beauty makeup foundation and seeks improvement and diverse attempts.

[0081] Beauty makeup steps represent each main operation link required to complete the initial recommended makeup look, such as steps of base makeup, eye makeup, lip makeup, blush, contouring, etc. Beauty makeup breakdown steps are the further refined operation processes of beauty makeup steps. For example, the base makeup step can be decomposed into more detailed steps such as taking foundation, application technique, and concealer position.

[0082] Exemplarily, collect the user's beauty tool purchase records, beauty video browsing history, search keywords, etc. on the beauty platform. Build a decision tree model based on the beauty tool purchase records, beauty video browsing history, and search keywords, and determine the user's beauty experience level based on the decision tree model. For example, if the user purchases basic beauty products and frequently browses beginner tutorials, the decision tree model tends to determine that the user is a beginner in beauty.

[0083] Build a beauty step knowledge base to store various common makeup looks and their corresponding standard steps. Extract the corresponding beauty steps from the knowledge base according to the type of the initial recommended makeup look.

[0084] When the user is a beginner in beauty, determine the beauty assistance mode as the learning mode. At this time, a GPT model based on the Transformer architecture can be used, and the GPT model is trained with a large amount of beauty data and user information to obtain a beauty large model. For each beauty step, set a detailed text template, and the beauty large model can generate specific text-based makeup guidance information according to the text template and the information related to the beauty step. For example, for the breakdown step of "pick up foundation", the beauty large model generates "pick up one pump of liquid foundation and use a beauty blender to dip the liquid foundation in small amounts multiple times" according to the text template.

[0085] Use a generative adversarial network to build a dataset containing various beauty operation images. The generative adversarial network selects appropriate elements from the dataset and synthesizes image-based makeup guidance information according to the description of the beauty breakdown steps, such as an image showing the specific action of a beauty blender dipping liquid foundation.

[0086] When the user is an advanced beauty user, determine the beauty assistance mode as the normal mode. Screen key images corresponding to the beauty steps from the image library, and the image library can be built by annotating and classifying a large number of beauty tutorial images. Use a content-based image retrieval algorithm to quickly retrieve the most representative image-based makeup guidance information from the image library according to the keywords of the beauty steps.

[0087] This embodiment provides a learning mode for beauty beginners, which helps users systematically learn makeup skills through detailed beauty breakdown steps, comprehensive text, and rich image guidance. The normal mode designed for advanced beauty users meets their need to quickly obtain key information with concise image guidance, saves time, and helps them try new makeup looks to achieve diverse improvements. The two modes accurately match the different needs of users, effectively improving the users' beauty learning and practice experience, and meeting the pursuit of beauty by users at all stages.

[0088] In an embodiment of the present disclosure, a beauty assistance method based on an artificial intelligence large model further includes: In response to receiving a virtual makeup try-on instruction: Generate a virtual makeup try-on effect based on the target recommended makeup and the third-person face information. The third-person face information is the user's face information collected during the virtual makeup try-on process.

[0089] Adjust the target recommended makeup based on the beauty makeup large model and the second user feedback information to obtain the target virtual makeup.

[0090] Generate a decomposed virtual makeup try-on effect based on the beauty makeup large model, the target virtual makeup, and the third-person face information.

[0091] Adjust the second beauty makeup guidance information based on the decomposed virtual makeup try-on effect to obtain the target beauty makeup guidance information.

[0092] In this embodiment, the virtual makeup try-on instruction refers to the operation instruction for the user to start the virtual makeup try-on function. For example, the virtual makeup try-on function is enabled by keyboard / voice input or by clicking "Virtual Makeup Try-On". The third-person face information includes the user's face information collected in real time during the virtual makeup try-on process, which is used to generate a virtual makeup try-on effect that is more suitable for the user in real time. The virtual makeup try-on effect refers to the try-on effect presented by simulating the application of the target recommended makeup to the third-person face information.

[0093] The beauty makeup large model is a large model obtained by training a GPT model based on the Transformer architecture by combining a large amount of beauty makeup data and user information. The second user feedback information includes the feedback information given by the user during the virtual makeup try-on process, which is used to adjust the target recommended makeup. The target virtual makeup is the virtual makeup obtained by adjusting the target recommended makeup using the beauty makeup large model according to the user's feedback.

[0094] The decomposed virtual makeup try-on effect decomposes the process of virtual makeup try-on and shows the effects of each step or stage. The decomposed virtual makeup try-on effect can be the virtual makeup try-on effect presented according to the makeup steps, such as first showing the effect of applying the base makeup and then showing the effect after adding eye makeup, etc. The target beauty makeup guidance information is the beauty makeup guidance information obtained by adjusting the second beauty makeup guidance information according to the target virtual makeup.

[0095] Exemplarily, using 3D face modeling technology and texture mapping technology, map the target recommended makeup information to the 3D model corresponding to the third-person face information to generate a virtual makeup try-on effect.

[0096] Use the BERT model to perform natural language processing on the second user feedback information to extract the user's intention. Input the user's intention into the beauty makeup large model to adjust the target recommended makeup to obtain the target virtual makeup. By simulating the makeup painting process and using animation technology, gradually apply the target virtual makeup to the third-person face information to generate a decomposed virtual makeup try-on effect.

[0097] According to the decomposed virtual makeup effect, update the text part of the second makeup guidance information using the makeup large model, and update the image part using image editing technology to obtain the target makeup guidance information.

[0098] This embodiment enables users to intuitively see the presentation effect of the target recommended makeup on their own faces, which can effectively reduce the makeup trial time and cost, reduce the number of touch-ups, and improve the makeup effect. Users can provide timely feedback based on the makeup trial effect, and a target virtual makeup that better meets their needs can be generated in real time. At the same time, the target makeup guidance information adjusted based on the decomposed makeup trial effect can provide more targeted makeup guidance for users and comprehensively improve the users' makeup experience.

[0099] In an embodiment of the present disclosure, determining an initial recommended makeup based on a user profile, a scene type, and first face information includes: In response to receiving an input instruction from a user, determine the scene type based on the input instruction, and determine a target recommended makeup type from multiple recommended makeup types based on the scene type.

[0100] Determine a target makeup dataset from multiple makeup datasets based on the target recommended makeup type.

[0101] Determine an initial recommended makeup from the target makeup dataset based on the user profile and the first face information.

[0102] In this embodiment, determining an initial recommended makeup from the target makeup dataset based on the user profile and the first face information includes: Determine face features and makeup features based on the target makeup dataset.

[0103] Cluster the target makeup dataset based on a clustering algorithm and the face features to obtain multiple first makeup data subsets.

[0104] Cluster the target makeup dataset based on a clustering algorithm and the makeup features to obtain multiple second makeup data subsets.

[0105] Determine the user's facial features based on the user profile and the first face information.

[0106] Determine the initial recommended makeup based on the user's facial features, multiple first makeup data subsets, and multiple second makeup data subsets.

[0107] In this embodiment, the user's input instruction can be voice information collected by a microphone on a makeup mirror. The user's input instruction can include a limitation on the scene type, such as specific scene descriptions like "daily commute" and "dinner party". The speech recognition large model in the makeup large model can recognize this input instruction.

[0108] The recommended makeup types are different pre-set categories of makeup, such as natural style, retro style, fashionable style, etc. Each recommended makeup type is set with applicable scene tags. The target recommended makeup type is the makeup type suitable for the current scene determined according to the user input instruction.

[0109] The makeup dataset is a collection that stores a large amount of makeup data, including the detailed information of each makeup, as well as the corresponding facial features and makeup features. The target makeup dataset is the makeup dataset that matches the target recommended makeup type. Each recommended makeup type corresponds to a makeup dataset.

[0110] Facial features include data that describe the physiological features of the face, such as face shape, facial feature proportions, facial contour, etc., and are used for cluster analysis of the makeup dataset. Makeup features are data that describe the characteristics of the makeup itself, including makeup style, color matching, makeup difficulty, etc.

[0111] The first makeup data subset is the makeup data subset obtained by clustering based on facial features. The second makeup data subset is the makeup data subset obtained by clustering based on makeup features.

[0112] Exemplarily, a mapping table between scene types and recommended makeup types is established and stored in the database. When receiving a user input instruction, the scene type is extracted through string matching or natural language processing technology, and the corresponding target recommended makeup type is queried in the mapping table.

[0113] An index is established for each recommended makeup type in the database. According to the index of the target recommended makeup type, the target makeup dataset is quickly located and extracted.

[0114] Using the K-Means clustering algorithm, the facial features in the target makeup dataset are transformed into feature vectors. A suitable K value is set (determined based on experience or through experiments), and the K-Means algorithm is run to cluster the makeup with similar facial features into multiple first makeup data subsets.

[0115] Using the K-Means clustering algorithm, the makeup features are transformed into feature vectors. A suitable K value is set, and multiple second makeup data subsets are obtained by running the algorithm.

[0116] Facial features are extracted from the user portrait and the first facial information. The face feature point detector in the Dlib library can be used to extract features such as face shape and facial feature proportions.

[0117] Calculate the cosine similarity between the user's facial features and the central feature vectors of each first makeup data subset. Set a similarity threshold, and filter out the first makeup data subsets with similarity higher than the threshold.

[0118] Calculate the cosine similarity between the user's facial features and the central feature vectors of each second makeup data subset, and filter out the second makeup data subsets with a similarity higher than the threshold.

[0119] From the data subsets that meet the above two conditions simultaneously, select one or more makeup looks as the initial recommended makeup looks according to the preset priority rules (such as giving priority to the makeup looks that best match the user's style preferences).

[0120] The method for determining the initial recommended makeup looks based on the user portrait, scenario, and facial information in this embodiment can significantly improve the user experience. By accurately identifying the user's instructions to determine the scenario and matching the target recommended makeup look type, it ensures that the recommendation fits the usage scenario. Using the clustering algorithm and feature matching, it fully considers the user's facial and makeup preferences and provides more accurate personalized recommendations. The whole process is efficient and intelligent, quickly finding the initial makeup look suitable for the user, saving time and enhancing satisfaction.

[0121] In an embodiment of the present disclosure, determining the initial recommended makeup looks based on the user's facial features, multiple first makeup data subsets, and multiple second makeup data subsets includes: Calculate the first matching degree between the user's facial features and all first makeup data subsets. For each first makeup data subset: if the first matching degree between the user's facial features and this first makeup data subset is greater than or equal to the first threshold, add this first makeup data subset to the first matching makeup look set.

[0122] Calculate the second matching degree between the user's facial features and all second makeup data subsets. For each second makeup data subset: if the second matching degree between the user's facial features and this second makeup data subset is greater than or equal to the second threshold, add this second makeup data subset to the second matching makeup look set.

[0123] Determine the initial recommended makeup looks based on the first matching makeup look set and the second matching makeup look set.

[0124] In this embodiment, calculating the first matching degree between the user's facial features and all first makeup data subsets includes: Calculate the first mean vector of all data points in the first makeup data subset.

[0125] Calculate the cluster center vector corresponding to the cluster center data point in the first makeup data subset.

[0126] Calculate the average vector of the cluster center vector and the first mean vector.

[0127] Calculate the first matching degree between the user's facial features and the average vector.

[0128] In this embodiment, the first makeup data subset contains multiple data points, and each data point represents a feature vector related to makeup. The first makeup data subset is a cluster formed after clustering. The cluster center vector is the data point at the central position within the cluster, which can better represent the central tendency of the data within the cluster.

[0129] The first mean vector is the average of the feature vectors of all data points in the first makeup data subset, and is used to summarize the overall feature trend of the data in this subset.

[0130] The average vector is the average of the cluster center vector and the first mean vector, which combines the overall features of the first makeup data subset and the typical features within the cluster.

[0131] The first matching degree is used to measure the similarity between the user's facial features and the average vector, so as to evaluate the matching situation between the user's facial features and the first makeup data subset.

[0132] In this embodiment, determining the initial recommended makeup based on the first matching makeup set and the second matching makeup set includes: Calculate the intersection data points of the first matching makeup set and the second matching makeup set. The intersection data points are the data points that are both included in the first matching makeup set and the second matching makeup set.

[0133] If the intersection data points are not empty, then use these intersection data points as the initial recommended makeup.

[0134] If the intersection data points are empty, then calculate the third matching degree between the user's facial features and all data points in the first matching makeup set and the second matching makeup set based on the weight coefficient; Sort all data points from largest to smallest based on the third matching degree of each data point to obtain a recommended makeup sequence, and use the recommended makeup sequence as the initial recommended makeup.

[0135] In this embodiment, calculating the third matching degree between the user's facial features and all data points in the first matching makeup set and the second matching makeup set based on the weight coefficient includes: Calculate the third matching degree between the user's facial features and all data points in the first matching makeup set and the second matching makeup set based on the matching degree formula and the weight coefficient.

[0136] The matching degree formula is:

[0137] Where, represents the third matching degree, represents the weight coefficient of the data points in the first matching makeup set, represents the weight coefficient of the data points in the second matching makeup set, Represents the first matching degree between the data points in the first set of matching makeup and the user's facial features. Represents the second matching degree between the data points in the second set of matching makeup and the user's facial features. Represents the feature vector corresponding to the data points in the first set of matching makeup. Represents the feature vector corresponding to the data points in the second set of matching makeup. Represents the feature vector corresponding to the user's facial features.

[0138] Exemplarily, Xiaomei is about to attend a dinner. The beauty makeup large model constructs her user portrait through Xiaomei's user information and also collects her current facial information. After receiving the scene instruction of "dinner" from Xiaomei, the beauty makeup large model determines the target recommended makeup type and filters out the target makeup dataset from multiple makeup datasets.

[0139] For the target makeup dataset, the beauty makeup large model clusters based on facial features to obtain the first subset of makeup data. There are multiple feature vector data points representing different dinner makeup looks in this subset. Through calculation, the first mean vector is [0.7, 0.5, 0.8, 0.6], the cluster center vector is [0.6, 0.6, 0.7, 0.7], and the average vector is [0.65, 0.55, 0.75, 0.65]. Xiaomei's facial feature vector is [0.6, 0.5, 0.8, 0.6]. After calculation, the first matching degree is relatively high, and this first subset of makeup data is included in the first set of matching makeup.

[0140] Meanwhile, a second set of matching makeup is obtained based on makeup feature clustering. The beauty makeup large model calculates the intersection data points of the two subsets and finds that the intersection data points are empty. At this time, the weight coefficient of the first set of matching makeup is set to 0.55, and the weight coefficient of the second set of matching makeup is set to 0.45. For all the data points in the two sets, the third matching degree is calculated according to the matching degree formula. For example, if the first matching degree of a certain data point in the first set of matching makeup is 0.8 and the second matching degree of a certain data point in the second set of matching makeup is 0.7, the third matching degree of this data point is calculated to be 0.76. Finally, sorted by the third matching degree, a dinner makeup sequence is recommended for Xiaomei to help her create a perfect makeup suitable for the dinner.

[0141] This embodiment can accurately match the scene, quickly locate the suitable target recommended makeup type according to the instruction input by the user, and ensure that the recommendation highly fits the usage scenario. Through the clustering algorithm and feature matching, it deeply considers the user's facial features and makeup preferences, providing strong support for personalized recommendation. When the intersection of the two sets of matching makeup is empty, the unique weight coefficient and matching degree formula can make a comprehensive trade-off to obtain a scientific and reasonable recommendation sequence. In summary, this embodiment improves the accuracy and efficiency of the recommendation, saves the user's time, helps the user easily find the desired makeup, and greatly enhances the user's beauty makeup experience.

[0142] A beauty makeup assistance method based on an artificial intelligence large model corresponding to the above embodiment Figure 5 The block diagram of a beauty makeup assistance system based on an artificial intelligence large model provided by an embodiment of the present disclosure. For ease of illustration, only parts related to the embodiments of the present disclosure are shown. Refer to Figure 5 The beauty makeup assistance system 20 based on an artificial intelligence large model includes: a user data analysis module 21, a makeup recommendation module 22, and a feedback optimization module 23.

[0143] Among them, the user data analysis module 21 is used to determine a user profile based on multi-source user information and determine a beauty makeup assistance mode based on the user profile.

[0144] The makeup recommendation module 22 is used to determine an initial recommended makeup based on the user profile, the scene type, and the first face information, and generate first beauty makeup guidance information based on the beauty makeup assistance mode and the initial recommended makeup. The first face information is the face information of the user before applying makeup.

[0145] The feedback optimization module 23 is used to adjust the initial recommended makeup based on the beauty makeup large model, the second face information, and the first user feedback information to obtain a target recommended makeup. Adjust the first beauty makeup guidance information based on the target recommended makeup to obtain second beauty makeup guidance information. The second face information is the face information of the user after applying makeup.

[0146] In an embodiment of the present disclosure, the beauty makeup assistance mode includes a learning mode and a normal mode. The user data analysis module 21 is specifically used to determine the user's beauty makeup experience level based on the user profile.

[0147] If the user's beauty makeup experience level is a beginner in beauty makeup, the beauty makeup assistance mode is the learning mode.

[0148] If the user's beauty makeup experience level is an advanced beauty makeup user, the beauty makeup assistance mode is the normal mode.

[0149] In an embodiment of the present disclosure, the first beauty makeup guidance information includes text-based makeup application guidance information and image-based makeup application guidance information. The makeup recommendation module 22 is specifically used to determine multiple makeup steps based on the initial recommended makeup.

[0150] If the beauty makeup assistance mode is the learning mode: Generate multiple makeup breakdown steps based on each makeup step. Generate text-based makeup application guidance information and image-based makeup application guidance information based on each makeup breakdown step.

[0151] If the beauty makeup assistance mode is the normal mode: Generate image-based makeup application guidance information based on each makeup step.

[0152] In one embodiment of the present disclosure, a beauty makeup assistance system 20 based on an artificial intelligence large model further includes: a virtual makeup try-on module, configured to respond to receiving a virtual makeup try-on instruction: Generate a virtual makeup try-on effect based on a target recommended makeup look and third-person face information. The third-person face information is user face information collected during the virtual makeup try-on process.

[0153] Adjust the target recommended makeup look based on the beauty makeup large model and the second user feedback information to obtain a target virtual makeup look.

[0154] Generate a decomposed virtual makeup try-on effect based on the beauty makeup large model, the target virtual makeup look, and the third-person face information.

[0155] Adjust the second beauty makeup guidance information based on the decomposed virtual makeup try-on effect to obtain target beauty makeup guidance information.

[0156] In one embodiment of the present disclosure, the makeup look recommendation module 22 is specifically further configured to respond to receiving a user input instruction, determine a scene type based on the input instruction, and determine a target recommended makeup look type from multiple recommended makeup look types based on the scene type.

[0157] Determine a target makeup look dataset from multiple makeup look datasets based on the target recommended makeup look type.

[0158] Determine an initial recommended makeup look from the target makeup look dataset based on the user portrait and the first face information.

[0159] In one embodiment of the present disclosure, the makeup look recommendation module 22 is specifically further configured to determine face features and makeup look features based on the target makeup look dataset.

[0160] Cluster the target makeup look dataset based on a clustering algorithm and the face features to obtain multiple first makeup look data subsets.

[0161] Cluster the target makeup look dataset based on a clustering algorithm and the makeup look features to obtain multiple second makeup look data subsets.

[0162] Determine user facial features based on the user portrait and the first face information.

[0163] Determine an initial recommended makeup look based on the user facial features, the multiple first makeup look data subsets, and the multiple second makeup look data subsets.

[0164] In one embodiment of the present disclosure, the makeup look recommendation module 22 is specifically further configured to calculate a first matching degree between the user facial features and all the first makeup look data subsets. For each first makeup look data subset: if the first matching degree between the user facial features and the first makeup look data subset is greater than or equal to a first threshold, add the first makeup look data subset to the first matching makeup look set.

[0165] Calculate the second matching degree between the user's facial features and all second makeup data subsets. For each second makeup data subset: If the second matching degree between the user's facial features and the second makeup data subset is greater than or equal to the second threshold, add the second makeup data subset to the second matching makeup set.

[0166] Determine the initial recommended makeup based on the first matching makeup set and the second matching makeup set.

[0167] See Figure 6 , Figure 6 is a schematic block diagram of an electronic device provided by an embodiment of the present disclosure. As Figure 6 shown, the electronic device 300 in this embodiment may include: one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The above-mentioned processors 301, input devices 302, output devices 303, and memories 304 communicate with each other through a communication bus 305. The memory 304 is used to store computer programs, and the computer programs include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. Among them, the processor 301 is configured to call the program instructions to execute the functions of each module in the above system embodiments, such as Figure 5 the functions of the modules 21 to 23 shown.

[0168] It should be understood that in the embodiments of the present disclosure, the so-called processor 301 may be a central processing unit (Central Processing Unit, CPU), and this processor may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), field programmable gate arrays (Field-Programmable Gate Array, FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0169] The input device 302 may include a touchpad, a fingerprint acquisition sensor (for acquiring the user's fingerprint information and the direction information of the fingerprint), a microphone, etc., and the output device 303 may include a display (such as LCD), a speaker, etc.

[0170] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A part of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.

[0171] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure may execute the implementation manners described in the first and second embodiments of a beauty makeup assistance method based on an artificial intelligence large model provided by the embodiments of the present disclosure, and may also execute the implementation manner of the electronic device 300 described in the embodiments of the present disclosure, which will not be elaborated herein.

[0172] In another embodiment of the present disclosure, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, all or part of the processes in the methods of the above embodiments are implemented. It can also be completed by instructing relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0173] The computer-readable storage medium may be an internal storage unit of the electronic device in any of the foregoing embodiments, such as the hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the electronic device. Further, the computer-readable storage medium may also include both the internal storage unit and the external storage device of the electronic device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store the data that has been output or will be output.

[0174] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this disclosure.

[0175] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described electronic devices and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0176] In several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed couplings, direct couplings, or communication connections to each other can be indirect couplings or communication connections through some interfaces or units, or can be electrical, mechanical, or other forms of connection.

[0177] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of this disclosure.

[0178] In addition, the functional units in each embodiment of this disclosure can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0179] The above is only the specific implementation manner of this disclosure, but the protection scope of this disclosure is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or substitutions within the technical scope disclosed by this disclosure, and these modifications or substitutions should all be covered by the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be subject to the protection scope of the claims.

Claims

1. A beauty makeup assistance method based on an artificial intelligence large model, characterized in that: include: Determine a user portrait based on multi-source user information, and determine a beauty assistance mode based on the user portrait; Determine an initial recommended makeup based on the user portrait, the scene type and the first facial information, and generate first makeup guidance information based on the makeup assistance mode and the initial recommended makeup; the first facial information is facial information of the user before putting on makeup; The initial recommended makeup is adjusted based on the beauty model, the second facial information and the first user feedback information to obtain the target recommended makeup; the first beauty guidance information is adjusted based on the target recommended makeup to obtain the second beauty guidance information; the second facial information is the facial information of the user after applying makeup.

2. A beauty makeup assisting method based on artificial intelligence big model as claimed in claim 1, characterized in that: The beauty assistance mode includes a learning mode and a normal mode; The determining of the beauty assistance mode based on the user portrait includes: Determining the user's beauty experience level based on the user portrait; If the user's makeup experience level is entry-level, the makeup assistance mode is the learning mode; If the user's beauty experience level is advanced beauty, the beauty assistance mode is normal mode.

3. A beauty makeup assisting method based on artificial intelligence big model as claimed in claim 2, characterized in that: The first makeup guidance information includes text-based makeup guidance information and image-based makeup guidance information; The generating first beauty makeup guidance information based on the beauty makeup assistance mode and the initial recommended makeup includes: Determining a plurality of makeup steps based on the initial recommended makeup; If the makeup assistance mode is the learning mode: Generate multiple makeup decomposition steps based on each makeup step; generate text-based makeup guidance information and image-based makeup guidance information based on each makeup decomposition step; If the makeup assistance mode is the normal mode: Generate image-based makeup guidance information based on each makeup step.

4. The beauty makeup assisting method based on artificial intelligence big model as claimed in claim 1, characterized in that: Also includes: In response to receiving virtual try-on instructions: Generate a virtual makeup trial effect based on the target recommended makeup and third-person face information; the third-person face information is user face information collected during the virtual makeup trial process; Adjusting the target recommended makeup based on the beauty makeup model and the second user feedback information to obtain a target virtual makeup; Generate a decomposed virtual makeup trial effect based on the beauty makeup model, the target virtual makeup and the third person's face information; The second beauty makeup guidance information is adjusted based on the decomposed virtual makeup trial effect to obtain target beauty makeup guidance information.

5. The beauty makeup assisting method based on artificial intelligence big model as claimed in claim 1, characterized in that: The determining of the initial recommended makeup based on the user portrait, the scene type and the first face information includes: In response to receiving an input instruction from a user, determining a scene type based on the input instruction, and determining a target recommended makeup type from a plurality of recommended makeup types based on the scene type; Determining a target makeup dataset from multiple makeup datasets based on the target recommended makeup type; An initial recommended makeup is determined from the target makeup dataset based on the user portrait and the first face information.

6. A beauty makeup assisting method based on artificial intelligence big model as claimed in claim 5, characterized in that: The determining an initial recommended makeup from the target makeup dataset based on the user portrait and the first face information includes: Determine facial features and makeup features based on a target makeup dataset; Clustering the target makeup data set based on a clustering algorithm and facial features to obtain multiple first makeup data subsets; Clustering the target makeup data set based on a clustering algorithm and makeup features to obtain multiple second makeup data subsets; Determine the user's facial features based on the user portrait and the first face information; An initial recommended makeup is determined based on the user facial features, the plurality of first makeup data subsets, and the plurality of second makeup data subsets.

7. A beauty makeup assisting method based on artificial intelligence big model as claimed in claim 6, characterized in that: The determining of an initial recommended makeup based on the user's facial features, the plurality of first makeup data subsets, and the plurality of second makeup data subsets comprises: Calculating a first matching degree between the user's facial features and all first makeup data subsets; for each first makeup data subset: if the first matching degree between the user's facial features and the first makeup data subset is greater than or equal to a first threshold, adding the first makeup data subset to a first matching makeup set; Calculating a second matching degree between the user's facial features and all second makeup data subsets; for each second makeup data subset: if the second matching degree between the user's facial features and the second makeup data subset is greater than or equal to a second threshold, adding the second makeup data subset to a second matching makeup set; An initial recommended makeup is determined based on the first matching makeup set and the second matching makeup set.

8. A beauty makeup assistance system based on an artificial intelligence big model, characterized in that: include: A user data analysis module, used to determine a user portrait based on multi-source user information, and determine a beauty assistance mode based on the user portrait; a makeup recommendation module, configured to determine an initial recommended makeup based on the user portrait, the scene type and the first facial information, and generate first makeup guidance information based on the makeup assistance mode and the initial recommended makeup; the first facial information is facial information of the user before putting on makeup; A feedback optimization module is used to adjust the initial recommended makeup based on the beauty model, the second facial information and the first user feedback information to obtain a target recommended makeup; based on the target recommended makeup, the first beauty guidance information is adjusted to obtain the second beauty guidance information; the second facial information is the facial information of the user after applying makeup.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images