Cosmetic product recommendation method and device based on skin identification, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence, and provides a beauty makeup product recommendation method and device based on skin identification, equipment and a medium, which can call a multi-modal feature extraction engine to extract multi-modal features of a target facial image, and call a skin identification model to process the multi-modal features to obtain a target skin identification result. The skin condition can be accurately identified based on multi-modal features; inputting the target skin identification result into an inference engine loaded with an inference rule base to obtain a user demand tag list sequenced according to tag weights, so that the credibility of inference can be improved based on professional knowledge; the user demand tag list is converted into the query vector, and the multi-path recall strategy engine is called to execute recall in the commodity feature vector library based on the query vector, so that the comprehensiveness and accuracy of commodity recall are improved; and the candidate makeup product set is optimized based on a multiple optimization constraint mechanism to obtain the recommended makeup product list, so that the accuracy of the recommended makeup products can be further improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, and medium for recommending beauty products based on skin type recognition. Background Technology

[0002] In existing technologies, skincare product recommendations on beauty e-commerce platforms primarily rely on users' historical behavioral data (such as browsing and purchase records) or static information (such as skin type and age), and are made through rule matching or collaborative filtering. While these methods are effective in the cold start phase, they lack specificity and fail to meet users' growing personalized skincare needs because they cannot perceive the user's current skin condition.

[0003] Existing technologies also attempt to incorporate image recognition technology to analyze user-uploaded facial images to determine basic information such as skin texture and skin tone, and then recommend products accordingly. However, the following main problems still exist in the overall process: (1) Limited recognition dimensions: Image models cannot analyze local features in detail, such as oil in the T-zone (forehead and nose), large pores, sensitive areas, etc. (2) Lack of structured label output: The recognition results are usually a single label, which cannot form a standard structured skin state; (3) Images and products cannot be linked: Product information is mostly text or image descriptions, without unified semantic modeling, making it difficult to effectively match with skin analysis results; (4) The recommendation system does not integrate an image-driven entry point: the recommendation process is mainly based on user behavior and fails to realize the recommendation mechanism of "directly driving recall by image analysis results"; (5) Fragmented system and unclosed links: Image analysis and recommendation systems are deployed independently, data cannot be shared, and AI (Artificial Intelligence) capabilities cannot truly participate in the marketing process; (6) Skin quality testing is costly: it requires professional equipment and offline services, making it difficult to apply on a large scale in e-commerce scenarios. Summary of the Invention

[0004] In view of the above, it is necessary to provide a method, device, equipment and medium for recommending beauty products based on skin type recognition, in order to solve the problem of inaccurate recommendations caused by poor skin type recognition in the process of recommending beauty products.

[0005] A method for recommending beauty products based on skin type identification, the method comprising: In response to a beauty product recommendation instruction triggered based on a target user's facial image, a multimodal feature extraction engine is invoked to extract the multimodal features of the target facial image; The multimodal features are processed by calling the skin texture recognition model to obtain the target skin texture recognition result; Obtain a reasoning rule base built based on professional knowledge, and load the reasoning rule base into the reasoning engine; The target skin type recognition result is input into the inference engine to obtain a list of user demand tags sorted by tag weight; The user demand tag list is converted into a query vector, and the multi-way recall strategy engine is invoked to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products. The candidate beauty product set is optimized based on multiple optimization constraint mechanisms to obtain a recommended beauty product list.

[0006] A beauty product recommendation device based on skin type recognition, the beauty product recommendation device based on skin type recognition includes: The extraction unit is used to respond to a beauty product recommendation instruction triggered based on a target user's target facial image, and to call a multimodal feature extraction engine to extract the multimodal features of the target facial image; The recognition unit is used to call the skin texture recognition model to process the multimodal features and obtain the target skin texture recognition result; The loading unit is used to obtain a reasoning rule base built based on professional knowledge and load the reasoning rule base into the reasoning engine; The inference unit is used to input the target skin texture recognition result into the inference engine to obtain a list of user demand tags sorted according to tag weights. The recall unit is used to convert the user demand tag list into a query vector and call the multi-way recall strategy engine to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products. The optimization unit is used to optimize the candidate beauty product set based on multiple optimization constraint mechanisms to obtain a recommended beauty product list.

[0007] A computer device, the computer device comprising: Memory, storing at least one instruction; and The processor executes the instructions stored in the memory to implement the skin type recognition-based beauty product recommendation method.

[0008] A computer-readable storage medium storing at least one instruction, which is executed by a processor in a computer device to implement the skin type identification-based beauty product recommendation method.

[0009] As can be seen from the above technical solutions, this invention can call a multimodal feature extraction engine to extract multimodal features from a target facial image, and call a skin texture recognition model to process the multimodal features to obtain the target skin texture recognition result, which can accurately identify skin texture based on multimodal features; inputting the target skin texture recognition result into an inference engine loaded with an inference rule base, obtaining a list of user demand tags sorted by tag weight, which can improve the credibility of inference based on professional knowledge; converting the list of user demand tags into a query vector, and calling a multi-way recall strategy engine to perform recall in the product feature vector library based on the query vector, which improves the comprehensiveness and accuracy of product recall; optimizing the candidate beauty product set based on a multi-optimization constraint mechanism to obtain a recommended beauty product list, which can further improve the accuracy of recommended beauty products. Attached Figure Description

[0010] Figure 1 This is a flowchart of a preferred embodiment of the beauty product recommendation method based on skin type identification of the present invention; Figure 2 This is a functional block diagram of a preferred embodiment of the beauty product recommendation device based on skin type recognition of the present invention; Figure 3 This is a schematic diagram of the structure of a computer device that implements a preferred embodiment of the skin type-based beauty product recommendation method of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] like Figure 1 The diagram shown is a flowchart of a preferred embodiment of the beauty product recommendation method based on skin type recognition according to the present invention. The order of the steps in this flowchart can be changed, and some steps can be omitted, depending on different needs.

[0013] The skin texture-based beauty product recommendation method is applied to one or more computer devices. The computer device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0014] The computer device can be any electronic product that can interact with the user, such as a personal computer, tablet computer, smartphone, personal digital assistant (PDA), game console, interactive network television (IPTV), smart wearable device, etc.

[0015] The computer equipment may also include network equipment and / or user equipment. The network equipment includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0016] The server can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0017] Artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0018] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0019] The network in which the computer device is located includes, but is not limited to, the Internet, wide area network, metropolitan area network, local area network, and virtual private network (VPN).

[0020] S10, in response to a beauty product recommendation instruction triggered based on the target user's target facial image, the multimodal feature extraction engine is invoked to extract the multimodal features of the target facial image.

[0021] In this embodiment, the target user can be a user who needs to purchase beauty products.

[0022] In this embodiment, the target facial image can be uploaded by the target user.

[0023] In this embodiment, the beauty product recommendation instruction can be automatically triggered when the target facial image is detected to be uploaded to a designated platform.

[0024] In this embodiment, the step of calling the multimodal feature extraction engine to extract the multimodal features of the target facial image includes: Perform image quality detection on the target facial image; When the target facial image passes the image quality detection, the target facial image is preprocessed to obtain the image to be processed; The face positions in the image to be processed are located and standardized using a face detection and alignment algorithm. The skin region at the face location is segmented using a semantic segmentation model, and the region of interest (ROI) of the skin region is extracted. Extract the texture features, color features, geometric features, and deep learning features of the region of interest; The multimodal features are obtained by fusing the texture features, color features, geometric features, and deep learning features based on an attention mechanism.

[0025] The image quality of the target facial image can be determined by detecting its resolution, sharpness, and illumination.

[0026] The preprocessing of the target facial image includes performing image enhancement, noise reduction, and color correction on the target facial image in sequence.

[0027] By performing image quality inspection and preprocessing, the reliability of the analyzed data can be ensured.

[0028] Among them, MTCNN (Multi-task Cascaded Convolutional Networks) or RetinaFace (retinal face detection model) can be used to locate and normalize the face position in the image to be processed.

[0029] Specifically, the semantic segmentation model can be used to segment the skin region of the face from the background of the target facial image.

[0030] The region of interest may include key skin areas such as the T-zone (forehead and nose), cheeks, eye area, and lip area.

[0031] Specifically, an AI feature extraction engine can be invoked to extract multimodal features for skin texture. For example, texture features can be extracted using Local Binary Pattern (LBP), Gray Level Co-occurrence Matrix (GLCM), Gabor filtering (a filter used for texture analysis, etc.); color features can be extracted based on HSV (Hue-Saturation-Value) and Lab (uniform color space) color space analysis; pore density, wrinkle depth, and skin surface roughness can be extracted as geometric features; and high-level skin features can be obtained using convolutional neural network models such as EfficientNet, ResNet, and Vision Transformer as deep learning features.

[0032] In the above embodiments, multiple modal features extracted can be fused based on an attention mechanism, thereby assisting in the subsequent comprehensive assessment of the user's skin condition.

[0033] S11, the skin texture recognition model is invoked to process the multimodal features to obtain the target skin texture recognition result.

[0034] In this embodiment, the skin texture recognition model includes a skin texture classification model, a skin problem detection model, and a skin age prediction model.

[0035] In this embodiment, the step of calling the skin texture recognition model to process the multimodal features and obtain the target skin texture recognition result includes: The multimodal features are input into the skin type classification model to obtain the target skin type and the skin type confidence level corresponding to the target skin type; The multimodal features are input into the skin problem detection model to obtain the target skin problem and a severity score of the target skin problem; Extract skin aging features from the multimodal features, input the skin aging features into the skin age prediction model to obtain the predicted skin age, obtain the actual age of the target user, and perform skin aging trend analysis based on the predicted skin age and the actual age to obtain the skin aging trend analysis result; By integrating the target skin type and its corresponding confidence level, the target skin problem and its severity score, and the skin aging trend analysis results, the target skin type identification result is obtained.

[0036] The target skin type can be oily, dry, sensitive, or combination.

[0037] The skin type confidence score is used to reflect the confidence level of the identified skin type.

[0038] The target skin problems can include acne, blackheads, blemishes, wrinkles, enlarged pores, dull skin tone, etc.

[0039] Among these methods, techniques such as target detection and image analysis can be used to assess the presence or severity of skin problems as a severity score for the problem.

[0040] The skin age prediction model can be a deep learning regression model or other model with predictive capabilities.

[0041] All of the above analysis processes are conducted under the premise of ensuring user privacy and security. For example, the processing of image data must follow security protocols, be used only for real-time analysis and will not be improperly stored, thereby ensuring that users' personal information is protected.

[0042] The above embodiments enable a comprehensive analysis of a user's skin condition without the need for additional instruments, thereby improving the accuracy of identification and reducing identification costs.

[0043] This embodiment utilizes a computer vision AI model to automatically detect skin texture in user-provided facial images, directly acquiring the user's true skin characteristics and problems. This avoids the subjective bias and errors inherent in traditional methods that rely on users filling out skin texture questionnaires. Compared to manual question-and-answer methods (studies show that 63% of women misjudge their own skin type), AI image analysis can objectively identify skin problems such as wrinkles, pigmentation, and sebum secretion, improving the consistency and accuracy of skin texture assessment. Simultaneously, the AI ​​skin texture analysis model used in this embodiment extracts multi-dimensional skin feature information from images, including texture details (such as skin surface roughness and wrinkle texture), color distribution (such as skin tone evenness, redness, and pigmentation), and geometric shapes (such as facial contours and proportions, and volume changes in specific areas). By fusing and modeling these multi-modal features through deep learning, different skin types can be more precisely distinguished and subtle skin problems identified. Compared to traditional methods that only utilize single-dimensional features, the comprehensive analysis of multi-dimensional data significantly improves the accuracy of skin texture classification and problem detection, ensuring that subsequent recommendations better match the user's actual skin needs.

[0044] S12, Obtain the reasoning rule base built based on professional knowledge, and load the reasoning rule base into the reasoning engine.

[0045] In this embodiment, the reasoning rule base may include a dermatology medical knowledge base, a cosmetic ingredient database, a skincare rule base, a user profile rule base, and an environmental factor database.

[0046] The dermatology medical knowledge base can store professional knowledge in the field of dermatology, including the causes of common skin symptoms (such as acne, pigmentation, and sensitivity) (e.g., acne is caused by excessive sebum secretion, clogged hair follicles, bacterial infection, etc.), care principles (e.g., sensitive skin needs to avoid irritating ingredients and use gentle care methods), and the symptom-cause-treatment ternary mapping relationship (e.g., "acne - excessive sebum secretion - oil control care", "pigmentation - ultraviolet radiation - sun protection + whitening care"). The dermatology medical knowledge base can be stored in a structured data format.

[0047] The cosmetic ingredient database can store information on common cosmetic ingredients, including ingredient name, efficacy (e.g., hyaluronic acid has moisturizing effects, salicylic acid has oil-controlling and exfoliating effects), suitability (e.g., ceramides are suitable for sensitive skin, alcohol is not suitable for sensitive skin), contraindications (e.g., pregnant women should avoid retinoic acid ingredients), efficacy-suitability-contraindication mapping relationship (e.g., "hyaluronic acid - suitable for all skin types - no specific contraindications", "salicylic acid - suitable for oily and combination skin types - use with caution for sensitive skin"), etc., and establish a correlation table between ingredients and skin type and skin problems.

[0048] The skincare rule base is used to summarize general skincare process rules, including daytime skincare routines, nighttime skincare routines, skincare plans adjusted according to the season (such as focusing on oil control and sun protection in summer, and focusing on moisturizing and repair in winter), age-appropriate skincare rules (such as focusing on basic moisturizing and preventing early signs of aging for 20-30 years old, and focusing on anti-aging and repair for those over 30 years old), and the mapping relationship between morning and evening skincare, seasonal adjustments, and age-appropriateness, etc., and form structured rule entries.

[0049] The user profile rule base is built based on user profile information and can include skin characteristics and care priorities for different age groups (e.g., teenagers are prone to oily skin and acne, so the focus of care is oil control and acne removal; middle-aged and elderly people have dry skin and are prone to wrinkles, so the focus of care is moisturizing and anti-aging), and differences in skin type and care preferences between genders (e.g., men have more sebum secretion and prefer refreshing products; women pay more attention to whitening and anti-aging and have higher requirements for product texture).

[0050] The environmental factor database is used to store environmental factor data such as weather (temperature, humidity), ultraviolet intensity, and air quality, and to record the effects of different environmental factors on the skin.

[0051] Furthermore, the rules and relationships in the aforementioned dermatology medical knowledge base, cosmetic ingredient database, skincare rule base, user profile rule base, and environmental factor database are converted into a rule format that the inference engine can recognize, and loaded into the inference engine for use in subsequent inference.

[0052] S13, input the target skin texture recognition result into the inference engine to obtain a list of user demand tags sorted by tag weight.

[0053] In this embodiment, inputting the target skin texture recognition result into the inference engine to obtain a list of user demand tags sorted by tag weight includes: The target skin type identification results are used to perform matching queries in the inference rule base to obtain basic care need tags, functional need tags, product type preference tags, ingredient preference tags, and usage scenario rule tags; Establish a tag set based on the basic care needs tag, the functional needs tag, the product type preference tag, the ingredient preference tag, and the usage scenario rule tag; Calculate the Term Frequency-Inverse Document Frequency (TF-IDF) weight for each tag in the tag set; Obtain the expert weight and confidence level for each label; The tag weight of each tag is determined based on the term frequency-inverse document frequency weight and the expert weight of each tag; Each tag is structured; The tags obtained after structured processing are sorted in descending order of their weights, and each tag is labeled using the confidence level to obtain the user demand tag list.

[0054] The basic care needs tags can include oil control and refreshing, soothing and anti-inflammatory, and gentle skincare; the functional needs tags can include whitening and fading spots, and deep moisturizing; the product type preference tags can include moisturizing masks, and daytime sun protection; the ingredient preference tags can include salicylic acid, azelaic acid, and so on; and the usage scenario rules tags can include refreshing and moisturizing during the day, oil control and sun protection in summer, and moisturizing and anti-drying in winter.

[0055] The word frequency is used to reflect the number of times each tag is associated with a user's skin problems, skin texture characteristics, etc.

[0056] The inverse document frequency is used to reflect the general importance of each tag.

[0057] The expert weights can be configured by dermatologists, beauty and skincare experts, etc., based on the importance of each tag to a user's skin problems.

[0058] The confidence level can be configured based on the confidence level of the model involved.

[0059] Through the above embodiments, it is possible to infer users' personalized needs based on expert knowledge, thereby helping to improve the accuracy of subsequent beauty product recommendations.

[0060] S14, the user demand tag list is converted into a query vector, and the multi-way recall strategy engine is called to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products.

[0061] In this embodiment, the list of user demand tags can be converted into the query vector using deep learning embedding models such as sentence vector models.

[0062] In this embodiment, before invoking the multi-path recall strategy engine to perform recall in the product feature vector library based on the query vector, the method further includes: Collect beauty product information; wherein, the beauty product information includes text description information, product image information, and product attribute information; The textual description information is transformed into a semantic vector using a text embedding model. The image feature vector of the product image information is extracted using a convolutional neural network; The product attribute information is encoded to obtain an attribute feature vector; The semantic vector, the image feature vector, and the attribute feature vector are fused to obtain a fused vector; An approximate nearest neighbor search index is constructed for the fused vector to obtain the product feature vector library.

[0063] The text description information may include the name, category, active ingredients, and applicable skin type of the beauty product.

[0064] The product image information may include multi-angle images of the product.

[0065] The product attribute information may include price, brand, user reviews, etc.

[0066] After collecting the beauty product information, the information can be cleaned and standardized to improve data quality, reduce data redundancy, and facilitate subsequent processing.

[0067] Specifically, the Faiss (Facebook AI Similarity Search) index, which uses the HNSW (Hierarchical Navigable Small Worlds) or IVF (Inverted File) algorithm, can be used to construct an approximate nearest neighbor search index for the fused vector to improve retrieval efficiency.

[0068] In this embodiment, the invocation of the multi-path recall strategy engine to perform recall in the product feature vector library based on the query vector, resulting in a candidate beauty product set, includes: Using the query vector, an approximate nearest neighbor search recall is performed in the product feature vector library to obtain the first set of beauty products; Based on the historical behavior and preference data of similar users corresponding to the target user, collaborative filtering is performed in the product feature vector library to obtain a second set of beauty products; Obtain the content tags corresponding to the target user, and perform content retrieval in the product feature vector library based on the content tags to obtain a third set of beauty products; Obtain current trend data, and perform popular trend recall in the product feature vector library based on the current trend data to obtain the fourth set of beauty products; The candidate beauty product set is obtained by calculating the union of the first set of beauty products, the second set of beauty products, the third set of beauty products, and the fourth set of beauty products.

[0069] In the process of performing near nearest neighbor search recall, the top K candidate products that best match the user's needs can be found according to indicators such as cosine similarity to construct the first set of beauty products.

[0070] Among them, when conducting collaborative filtering recall, products frequently purchased by users with similar skin types can be recalled.

[0071] When conducting content recall, user tags can be directly matched with product tags or attributes, such as filtering out products labeled "suitable for sensitive skin" or limiting the scope by category.

[0072] When recalling products based on popular trends, one can consider currently popular products and trends, such as recalling related products that have recently received high praise and sales.

[0073] In the above embodiments, multiple recall methods can be used to conduct multi-channel recall, so that the candidate beauty product set can cover multiple dimensions, including newly discovered products that meet the user's skin needs, as well as products with good public reputation.

[0074] In this embodiment, a multi-channel parallel recall mechanism is constructed during the product candidate recall stage. It employs multiple strategies, including content-based matching (such as semantic matching using multimodal vector representations of product images and text descriptions), collaborative filtering recall (such as mining similar users and products based on user history), and trend-based recall, to fuse candidate results from various channels into a base recommendation set. This multi-path recall approach balances the coverage and relevance of recommendation results, uncovering users' potential interests and preferences from different perspectives and avoiding overlooking long-tail products that users might like. In particular, by using multimodal vectorization technology to co-encode product image features and text attributes for recall matching, a deep correspondence between skin characteristics and product efficacy is ensured, thereby generating a more comprehensive and highly personalized set of candidate beauty products.

[0075] S15, the candidate beauty product set is optimized based on a multi-factor optimization constraint mechanism to obtain a recommended beauty product list.

[0076] In this embodiment, the optimization of the candidate beauty product set based on a multiple optimization constraint mechanism to obtain a recommended beauty product list includes: Configure the product weight of each beauty product in the candidate beauty product set according to the reliability of each recall method, and sort the beauty products in the candidate beauty product set according to the product weight to obtain the first candidate beauty product sequence. The first candidate beauty product sequence is rearranged using a machine learning ranking model to obtain the second candidate beauty product sequence. The historical behavior data and real-time feedback data of the target user are obtained, and the second candidate beauty product sequence is fine-tuned based on the historical behavior data and real-time feedback data of the target user to obtain the third candidate beauty product sequence. The configured business rule data is obtained, and the business rule data is used to filter and correct the third candidate beauty product sequence to obtain the fourth candidate beauty product sequence; wherein, the business rule data includes inventory verification rules, price range filtering rules, and brand preference rules; A diversity control algorithm was used to impose multi-dimensional constraints on the fourth candidate beauty product sequence to obtain the fifth candidate beauty product sequence. The recommended beauty product list is obtained by supplementing the display information of each candidate beauty product in the fifth candidate beauty product sequence.

[0077] When configuring product weights, candidates can be appropriately weighted based on the reliability and relevance of each channel. For example, products recalled directly by skin texture vectors have strong relevance and can have relatively higher weights; popular recalls may have slightly lower relevance and thus lower weights.

[0078] The machine learning ranking model can be trained based on gradient boosting decision trees or deep learning ranking networks, and large-scale user interaction data can be used during training.

[0079] The machine learning ranking model can comprehensively consider multiple features to score each candidate product, including: the product's matching degree with the user's skin type tags, the product's own quality and reputation rating, the product's consistency with the user's historical preferences, and the product's novelty and diversity. Through predictive analysis using machine learning, the ranking model can output a comprehensive relevance score for each candidate and reorder the product list accordingly, ensuring that the products that best meet the user's individual needs and have the highest overall evaluation are placed first, thereby improving the accuracy and persuasiveness of the recommendation results.

[0080] After machine learning ranking, the results are further personalized and filtered according to business rules to suit specific user preferences and business constraints.

[0081] Specifically, if the target user has frequently purchased a certain brand or has a clear preference for a certain type of product, the ranking of the relevant product can be appropriately increased in the sorting; conversely, if the target user explicitly states that they dislike a certain ingredient or product type, their probability of appearing will be reduced. Thus, the second candidate beauty product sequence is fine-tuned based on the target user's historical behavior data and real-time feedback data, making the recommended list more in line with the user's implicit preferences.

[0082] Specifically, when using the business rule data to filter and correct the third candidate beauty product sequence, the inventory verification rule can be used to remove products that are currently out of stock, thereby ensuring that the recommended products are effective in real time; the price range filtering rule can be used to exclude products that are significantly beyond the user's acceptable price range based on the user's spending power or set budget range; and the brand preference rule can be used to filter out brands that the target user is not interested in or prioritize retaining brands that the target user prefers, based on the target user's brand loyalty or aversion.

[0083] Furthermore, by utilizing the aforementioned diversity control algorithm, diversity constraints can be introduced in dimensions such as category, brand, and price to ensure that the recommendation results cover different brands, categories, and price ranges. This avoids the final recommendation list being too concentrated on a single type of product, thereby increasing the richness of the recommendation list and the possibility of meeting users' unspoken needs.

[0084] Furthermore, you can add product image links, names, prices, product IDs (identifiers), brief descriptions, etc., to improve the information needed to display the products.

[0085] In the above embodiments, the filtering adjustments can make the final recommended beauty products more in line with the user's personality and meet basic business requirements.

[0086] In this embodiment, during the fine-ranking stage, the candidate set is optimized by comprehensively considering multiple indicators such as skin type relevance, user interest prediction, and result diversity. Specifically, the model uses the matching degree between products and user skin type tags as an important weight, while combining the predicted scores of users' historical preferences (such as click-through rate or purchase tendency) to make ranking decisions, and appropriately introduces result diversity constraints to avoid the recommendation list being too homogeneous. Through the above multi-objective fusion strategy, a balance between relevance and diversity is achieved, improving the attractiveness of the final recommendation results to users. Research shows that recommendations that incorporate diversity considerations can improve user satisfaction and novelty experience. This embodiment ensures that the recommended content is highly matched with users' skin type needs, while also increasing the richness of the content and enhancing user satisfaction.

[0087] In this embodiment, after obtaining the recommended beauty product list, an AI confidence assessment model can be used to evaluate the overall reliability of the recommendations. Based on the model's confidence in the user's skin type analysis and matching, a quantitative indicator of the reliability of the results (such as a prediction confidence score or uncertainty range) is given. If the model is found to lack confidence in certain analysis conclusions, the recommendation is marked as low confidence. Next, an anomaly detection model is used to automatically capture anomalies in the process, including anomalies in the input image (such as image blurring or distortion, or failure to properly recognize faces) or anomalies in the analysis results (such as unreasonable skin type judgments or conflicting tag combinations). Once an anomaly is detected, timely measures are taken, such as activating backup recommendation schemes or prompting users to provide clearer photos. If necessary, a manual review mechanism can also be triggered. In the manual review mode, difficult cases can be submitted to beauty and skincare experts for annotation and suggestions as a supplementary correction to the AI ​​decision. At the same time, user feedback can be continuously collected, such as obtaining user satisfaction ratings for the recommendation results, click or purchase behavior, and any explicit feedback (such as users marking their likes or dislikes for recommended products) through front-end interaction records. The aggregated feedback data can be used for continuous model optimization. For example, the model can be retrained or its parameters adjusted based on real user feedback, or online learning technology can be used to allow the model to gradually adapt to new data. This embodiment can also periodically combine A / B testing results and version iteration plans to compare the performance metrics (such as click-through rate, conversion rate, retention rate, etc.) of different algorithm versions to select the best and update the recommendation model.

[0088] This embodiment, based on confidence assessment and adaptive feedback optimization, enables closed-loop optimization of the recommendation process. A confidence score is attached to each generated recommendation result, and user behavior feedback (such as clicks, browsing time, and feedback ratings) is continuously monitored. When the model is found to be uncertain about certain user preferences or the recommendation results do not meet expectations, the feedback data triggers an online learning process to adaptively adjust and optimize model parameters and recall and ranking strategies. By tracking changes in user behavior in real time and dynamically updating the model, the recommendation engine can continuously iterate itself, improving its ability to capture user interests and preferences. This feedback loop ensures the reliability and timeliness of the recommendation results, allowing the model to continuously evolve with changes in the environment and user needs, thus maintaining excellent recommendation performance over the long term.

[0089] Through the closed-loop process described above, the skin type recognition-based beauty product recommendation algorithm used in this embodiment can continuously evolve. While ensuring the accuracy and reliability of the results, it can also provide users with increasingly accurate and thoughtful personalized beauty product recommendation services over time.

[0090] In this embodiment, when multiple recommendation algorithms exist, an A / B testing strategy can be used to compare the recommendation results of different algorithms. Without affecting user experience, some user requests are redirected to different algorithm flows, and statistical analysis is used to compare the recommendation effects of each version to continuously optimize the recommendation strategy. Finally, the recommendation results are returned to the user through the front end (e.g., via page display or application pop-up), while the tracking ID and related context data of this recommendation are recorded on the back end for easy follow-up effect tracking and result interpretation. The output result data can also be in a structured format, including a list of recommended items and their associated information, facilitating presentation on the client side and storage and analysis in the logs.

[0091] This embodiment integrates multimodal vectorization of products, AI analysis of user faces, medical expert knowledge reasoning, a hybrid recall recommendation engine, and confidence monitoring feedback to form an end-to-end collaborative intelligent recommendation chain. Starting with a user taking a photo, the system processes and decides through modules, ultimately outputting a product list tailored to the user's skin type needs. It continuously improves through feedback and learning. This solution fully demonstrates innovative integration in AI, representing a qualitative leap compared to traditional methods.

[0092] This embodiment also improves deployment performance and optimizes user experience. Firstly, regarding recommendation accuracy, multimodal AI analysis and hybrid recall strategies effectively improve product matching precision, making recommended content more tailored to users' skin concerns and personal preferences. Secondly, in terms of response efficiency, the optimized system architecture enables millisecond-level real-time recommendation responses, utilizing efficient vector retrieval and hierarchical filtering technologies to meet low-latency interaction requirements. Finally, regarding user satisfaction and business performance, because the recommendation results are more intelligent, accurate, and diverse, users experience a better user experience, leading to increased conversion and retention rates. Overall, this embodiment outperforms existing technologies in terms of accuracy, intelligence, response speed, and user satisfaction, effectively overcoming the shortcomings of existing technologies in terms of subjectivity, singularity, and staticity. While providing more accurate and intelligent beauty product recommendation services, it also improves system response efficiency and user experience satisfaction, demonstrating significant value for widespread application.

[0093] As can be seen from the above technical solutions, this invention can call a multimodal feature extraction engine to extract multimodal features from a target facial image, and call a skin texture recognition model to process the multimodal features to obtain the target skin texture recognition result, which can accurately identify skin texture based on multimodal features; inputting the target skin texture recognition result into an inference engine loaded with an inference rule base, obtaining a list of user demand tags sorted by tag weight, which can improve the credibility of inference based on professional knowledge; converting the list of user demand tags into a query vector, and calling a multi-way recall strategy engine to perform recall in the product feature vector library based on the query vector, which improves the comprehensiveness and accuracy of product recall; optimizing the candidate beauty product set based on a multi-optimization constraint mechanism to obtain a recommended beauty product list, which can further improve the accuracy of recommended beauty products.

[0094] like Figure 2 The diagram shown is a functional block diagram of a preferred embodiment of the beauty product recommendation device based on skin texture recognition of the present invention. The beauty product recommendation device 11 based on skin texture recognition includes an extraction unit 110, an identification unit 111, a loading unit 112, an inference unit 113, a recall unit 114, and an optimization unit 115. The module / unit referred to in this invention refers to a series of computer program segments that can be executed by a processor and perform a fixed function, and are stored in memory. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0095] The extraction unit 110 is used to call the multimodal feature extraction engine to extract the multimodal features of the target facial image in response to a beauty product recommendation instruction triggered based on the target user's target facial image. The recognition unit 111 is used to call the skin texture recognition model to process the multimodal features and obtain the target skin texture recognition result; The loading unit 112 is used to obtain a reasoning rule base built based on professional knowledge and load the reasoning rule base into the reasoning engine; The inference unit 113 is used to input the target skin type recognition result into the inference engine to obtain a list of user demand tags sorted according to tag weights; The recall unit 114 is used to convert the user demand tag list into a query vector and call the multi-way recall strategy engine to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products. The optimization unit 115 is used to optimize the candidate beauty product set based on a multi-optimization constraint mechanism to obtain a recommended beauty product list.

[0096] As can be seen from the above technical solutions, this invention can call a multimodal feature extraction engine to extract multimodal features from a target facial image, and call a skin texture recognition model to process the multimodal features to obtain the target skin texture recognition result, which can accurately identify skin texture based on multimodal features; inputting the target skin texture recognition result into an inference engine loaded with an inference rule base, obtaining a list of user demand tags sorted by tag weight, which can improve the credibility of inference based on professional knowledge; converting the list of user demand tags into a query vector, and calling a multi-way recall strategy engine to perform recall in the product feature vector library based on the query vector, which improves the comprehensiveness and accuracy of product recall; optimizing the candidate beauty product set based on a multi-optimization constraint mechanism to obtain a recommended beauty product list, which can further improve the accuracy of recommended beauty products.

[0097] like Figure 3 The diagram shown is a schematic representation of the computer device used in a preferred embodiment of the skin type-based beauty product recommendation method of the present invention.

[0098] The computer device 1 may include a memory 12, a processor 13, and a bus (the arrow in the figure represents the bus), and may also include a computer program stored in the memory 12 and executable on the processor 13, such as a beauty product recommendation program based on skin type recognition.

[0099] Those skilled in the art will understand that the schematic diagram is merely an example of computer device 1 and does not constitute a limitation on computer device 1. Computer device 1 can be either a bus topology or a star topology. Computer device 1 may also include more or fewer other hardware or software than shown in the diagram, or different component arrangements. For example, computer device 1 may also include input / output devices, network access devices, etc.

[0100] It should be noted that the computer device 1 described is merely an example. Other existing or future electronic products that are adaptable to this invention should also be included within the scope of protection of this invention and are incorporated herein by reference.

[0101] The memory 12 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the computer device 1, such as a portable hard drive of the computer device 1. In other embodiments, the memory 12 can be an external storage device of the computer device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the computer device 1. Furthermore, the memory 12 can include both internal and external storage units of the computer device 1. The memory 12 can be used not only to store application software and various types of data installed on the computer device 1, such as the code of a beauty product recommendation program based on skin type recognition, but also to temporarily store data that has been output or will be output.

[0102] In some embodiments, the processor 13 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 13 is the control unit of the computer device 1, connecting various components of the computer device 1 via various interfaces and lines. It executes programs or modules stored in the memory 12 (e.g., executing a beauty product recommendation program based on skin type recognition) and calls data stored in the memory 12 to perform various functions of the computer device 1 and process data.

[0103] The processor 13 executes the operating system of the computer device 1 and various installed applications. The processor 13 executes these applications to implement the steps in the various embodiments of the skin type-based beauty product recommendation method described above, for example... Figure 1 The steps are shown.

[0104] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present invention. The one or more modules / units may be a series of computer-readable instruction segments capable of performing specific functions, which describe the execution process of the computer program in the computer device 1. For example, the computer program may be divided into an extraction unit 110, an identification unit 111, a loading unit 112, a reasoning unit 113, a recall unit 114, and an optimization unit 115.

[0105] The integrated unit implemented as a software functional module described above can be stored in a computer-readable storage medium. This software functional module, stored in a storage medium, includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute portions of the skin texture-based beauty product recommendation method described in the various embodiments of this invention.

[0106] If the modules / units integrated in the computer device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware devices. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above.

[0107] The computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory, etc.

[0108] Furthermore, the computer-readable storage medium may primarily include a stored program area and a stored data area, wherein the stored program area may store the operating system, an application program required for at least one function, etc.; and the stored data area may store data created based on the use of blockchain nodes, etc.

[0109] The blockchain referred to in this invention is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0110] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, in... Figure 3 The bus is represented by only one straight line, but this does not mean that there is only one bus or one type of bus. The bus is configured to enable communication between the memory 12 and at least one processor 13, etc.

[0111] Although not shown, the computer device 1 may also include a power supply (such as a battery) to power various components. Preferably, the power supply can be logically connected to the at least one processor 13 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The computer device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0112] Furthermore, the computer device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is typically used to establish communication connections between the computer device 1 and other computer devices.

[0113] Optionally, the computer device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the computer device 1 and to display a visual user interface.

[0114] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0115] It will be understood by those skilled in the art that Figure 3 The structure shown does not constitute a limitation on the computer device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0116] Combination Figure 1 The memory 12 in the computer device 1 stores multiple instructions to implement a beauty product recommendation method based on skin type recognition, and the processor 13 can execute the multiple instructions to achieve the following: In response to a beauty product recommendation instruction triggered based on a target user's facial image, a multimodal feature extraction engine is invoked to extract the multimodal features of the target facial image; The multimodal features are processed by calling the skin texture recognition model to obtain the target skin texture recognition result; Obtain a reasoning rule base built based on professional knowledge, and load the reasoning rule base into the reasoning engine; The target skin type recognition result is input into the inference engine to obtain a list of user demand tags sorted by tag weight; The user demand tag list is converted into a query vector, and the multi-way recall strategy engine is invoked to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products. The candidate beauty product set is optimized based on multiple optimization constraint mechanisms to obtain a recommended beauty product list.

[0117] Specifically, the processor 13's implementation method for the above instructions can be found in [reference needed]. Figure 1 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0118] It should be noted that all data involved in this case was legally obtained. Software tools or components not belonging to this company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0119] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0120] This invention can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0121] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0122] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0123] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0124] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0125] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices described in this invention can also be implemented by a single unit or device through software or hardware. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for recommending beauty products based on skin type recognition, characterized in that, The skin type-based beauty product recommendation method includes: In response to a beauty product recommendation instruction triggered based on a target user's facial image, a multimodal feature extraction engine is invoked to extract the multimodal features of the target facial image; The multimodal features are processed by calling the skin texture recognition model to obtain the target skin texture recognition result; Obtain a reasoning rule base built based on professional knowledge, and load the reasoning rule base into the reasoning engine; The target skin type recognition result is input into the inference engine to obtain a list of user demand tags sorted by tag weight; The user demand tag list is converted into a query vector, and the multi-way recall strategy engine is invoked to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products. The candidate beauty product set is optimized based on multiple optimization constraint mechanisms to obtain a recommended beauty product list.

2. The method for recommending beauty products based on skin texture recognition as described in claim 1, characterized in that, The step of calling the multimodal feature extraction engine to extract multimodal features from the target facial image includes: Perform image quality detection on the target facial image; When the target facial image passes the image quality detection, the target facial image is preprocessed to obtain the image to be processed; The face positions in the image to be processed are located and standardized using a face detection and alignment algorithm. The skin region at the location of the face is segmented using a semantic segmentation model, and the region of interest of the skin region is extracted. Extract the texture features, color features, geometric features, and deep learning features of the region of interest; The multimodal features are obtained by fusing the texture features, color features, geometric features, and deep learning features based on an attention mechanism.

3. The method for recommending beauty products based on skin texture recognition as described in claim 1, characterized in that, The skin texture recognition model includes a skin texture classification model, a skin problem detection model, and a skin age prediction model; The process of calling the skin texture recognition model to process the multimodal features and obtain the target skin texture recognition result includes: The multimodal features are input into the skin type classification model to obtain the target skin type and the skin type confidence level corresponding to the target skin type; The multimodal features are input into the skin problem detection model to obtain the target skin problem and a severity score of the target skin problem; Extract skin aging features from the multimodal features, input the skin aging features into the skin age prediction model to obtain the predicted skin age, obtain the actual age of the target user, and perform skin aging trend analysis based on the predicted skin age and the actual age to obtain the skin aging trend analysis result; By integrating the target skin type and its corresponding confidence level, the target skin problem and its severity score, and the skin aging trend analysis results, the target skin type identification result is obtained.

4. The method for recommending beauty products based on skin texture recognition as described in claim 1, characterized in that, The reasoning rule base includes a dermatology medical knowledge base, a cosmetic ingredient database, a skincare rule base, a user profile rule base, and an environmental factor database; The step of inputting the target skin texture recognition result into the inference engine to obtain a list of user demand tags sorted by tag weight includes: The target skin type identification results are used to perform matching queries in the inference rule base to obtain basic care need tags, functional need tags, product type preference tags, ingredient preference tags, and usage scenario rule tags; Establish a tag set based on the basic care needs tag, the functional needs tag, the product type preference tag, the ingredient preference tag, and the usage scenario rule tag; Calculate the term frequency-inverse document frequency weight for each tag in the tag set; Obtain the expert weight and confidence level for each label; The tag weight of each tag is determined based on the term frequency-inverse document frequency weight and the expert weight of each tag; Each tag is structured; The tags obtained after structured processing are sorted in descending order of their weights, and each tag is labeled using the confidence level to obtain the user demand tag list.

5. The method for recommending beauty products based on skin texture recognition as described in claim 1, characterized in that, Before invoking the multi-path recall strategy engine to perform recall in the product feature vector library based on the query vector, the method further includes: Collect beauty product information; wherein, the beauty product information includes text description information, product image information, and product attribute information; The textual description information is transformed into a semantic vector using a text embedding model. The image feature vector of the product image information is extracted using a convolutional neural network; The product attribute information is encoded to obtain an attribute feature vector; The semantic vector, the image feature vector, and the attribute feature vector are fused to obtain a fused vector; An approximate nearest neighbor search index is constructed for the fused vector to obtain the product feature vector library.

6. The method for recommending beauty products based on skin texture recognition as described in claim 1, characterized in that, The multi-path recall strategy engine is invoked to perform recall in the product feature vector library based on the query vector, resulting in a candidate set of beauty products including: Using the query vector, an approximate nearest neighbor search recall is performed in the product feature vector library to obtain the first set of beauty products; Based on the historical behavior and preference data of similar users corresponding to the target user, collaborative filtering is performed in the product feature vector library to obtain a second set of beauty products; Obtain the content tags corresponding to the target user, and perform content retrieval in the product feature vector library based on the content tags to obtain a third set of beauty products; Obtain current trend data, and perform popular trend recall in the product feature vector library based on the current trend data to obtain the fourth set of beauty products; The candidate beauty product set is obtained by calculating the union of the first set of beauty products, the second set of beauty products, the third set of beauty products, and the fourth set of beauty products.

7. The method for recommending beauty products based on skin texture recognition as described in claim 1, characterized in that, The optimization of the candidate beauty product set based on a multiple optimization constraint mechanism yields a recommended beauty product list, including: Configure the product weight of each beauty product in the candidate beauty product set according to the reliability of each recall method, and sort the beauty products in the candidate beauty product set according to the product weight to obtain the first candidate beauty product sequence. The first candidate beauty product sequence is rearranged using a machine learning ranking model to obtain the second candidate beauty product sequence. The historical behavior data and real-time feedback data of the target user are obtained, and the second candidate beauty product sequence is fine-tuned based on the historical behavior data and real-time feedback data of the target user to obtain the third candidate beauty product sequence. The configured business rule data is obtained, and the business rule data is used to filter and correct the third candidate beauty product sequence to obtain the fourth candidate beauty product sequence; wherein, the business rule data includes inventory verification rules, price range filtering rules, and brand preference rules; A diversity control algorithm was used to impose multi-dimensional constraints on the fourth candidate beauty product sequence to obtain the fifth candidate beauty product sequence. The recommended beauty product list is obtained by supplementing the display information of each candidate beauty product in the fifth candidate beauty product sequence.

8. A beauty product recommendation device based on skin type recognition, characterized in that, The skin type recognition-based beauty product recommendation device includes: The extraction unit is used to respond to a beauty product recommendation instruction triggered based on a target user's target facial image, and to call a multimodal feature extraction engine to extract the multimodal features of the target facial image; The recognition unit is used to call the skin texture recognition model to process the multimodal features and obtain the target skin texture recognition result; The loading unit is used to obtain a reasoning rule base built based on professional knowledge and load the reasoning rule base into the reasoning engine; The inference unit is used to input the target skin texture recognition result into the inference engine to obtain a list of user demand tags sorted according to tag weights. The recall unit is used to convert the user demand tag list into a query vector and call the multi-way recall strategy engine to perform recall in the product feature vector library based on the query vector to obtain a set of candidate beauty products. The optimization unit is used to optimize the candidate beauty product set based on multiple optimization constraint mechanisms to obtain a recommended beauty product list.

9. A computer device, characterized in that, The computer device includes: Memory, storing at least one instruction; and The processor executes instructions stored in the memory to implement the skin type identification-based beauty product recommendation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, which is executed by a processor in a computer device to implement the skin type identification-based beauty product recommendation method as described in any one of claims 1 to 7.

Images