WO2020156109A1 - 产品信息处理方法、装置、计算机设备和存储介质 - Google Patents

产品信息处理方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2020156109A1
WO2020156109A1 PCT/CN2020/071631 CN2020071631W WO2020156109A1 WO 2020156109 A1 WO2020156109 A1 WO 2020156109A1 CN 2020071631 W CN2020071631 W CN 2020071631W WO 2020156109 A1 WO2020156109 A1 WO 2020156109A1
Authority
WO
WIPO (PCT)
Prior art keywords
attribute
value
skin
target user
dimension
Prior art date
Application number
PCT/CN2020/071631
Other languages
English (en)
French (fr)
Inventor
仇子龙
曾婧
沙列夫莉雅
陈思
吴陶钧
Original Assignee
深圳碳云智能数字生命健康管理有限公司
深圳数字生命研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳碳云智能数字生命健康管理有限公司, 深圳数字生命研究院 filed Critical 深圳碳云智能数字生命健康管理有限公司
Publication of WO2020156109A1 publication Critical patent/WO2020156109A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations

Definitions

  • This application relates to the field of computers, in particular to a product information processing method, device, computer equipment and storage medium.
  • the skin presents different states according to various conditions such as environment, diet, skin care, and personal factors of the human body.
  • people usually choose skin care products based on personal past use or recommendations from other users.
  • individuals usually choose products that they often use based on personal preferences.
  • Recommendations by other users are usually recommended by other users based on their own products.
  • the accuracy of the skin care products recommended by the two methods is low.
  • a product information processing method includes:
  • Attributes are used to characterize a product index attribute related to the skin
  • the total second dimension attribute of each skin care product in the product information set is determined. value
  • the skin care product to be pushed corresponding to the target user is determined according to the total value of the second dimension attribute of each skin care product in the product information set and preset pushing conditions.
  • the acquiring an image containing the skin of the target user, and processing the image to obtain at least one first dimensional attribute value corresponding to the skin of the target user includes:
  • the skin attribute evaluation model is based on the Dimensional attribute value image training.
  • the method for generating the skin attribute evaluation model includes:
  • the method for generating the skin attribute evaluation model includes:
  • the product information set in the product information set is determined according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product in the preset product information set After the total value of the second dimension attribute of each skin care product, the method further includes:
  • the total value of the second dimension attribute of each skin care product in the product information set is normalized.
  • the normalizing the total value of the second dimension attribute of each skin care product in the product information set includes:
  • the highest value of the total value of the second dimension attribute of each skin care product in the product information set corresponds to the upper limit of the preset range, and the second dimension attribute of each skin care product in the product information set is totaled
  • the lowest value of the value corresponds to the lower limit of the preset range, and then the total value between the highest value and the lowest value of the second dimension attribute total value of each skin care product in the product information set corresponds to the total value Within the preset range.
  • the at least one second dimensional attribute value corresponding to the target user is obtained by combining the correlation relationship between the first dimensional attribute and the second dimensional attribute according to the first dimensional attribute value, include:
  • each first dimensional attribute corresponding to each second dimensional attribute of the target user is obtained according to the value of each first dimensional attribute corresponding to the target user and the corresponding correlation coefficient.
  • Combination values including:
  • each first dimensional attribute corresponding to each second dimensional attribute of the target user according to the value of each first dimensional attribute corresponding to the target user, the weight coefficient of each first dimensional attribute, and the corresponding correlation coefficient The combined value of.
  • each second dimensional attribute of the target user is obtained according to each first dimensional attribute value of the target user, the weight coefficient of each first dimensional attribute, and the corresponding correlation coefficient
  • the corresponding combined value of each first dimension attribute includes:
  • the weight coefficient of the first dimensional attribute, and the corresponding correlation coefficient the first dimensional attribute corresponding to each second dimensional attribute of the target user is obtained The combined value of.
  • each first dimensional attribute corresponding to each second dimensional attribute of the target user is obtained according to the value of each first dimensional attribute corresponding to the target user and the corresponding correlation coefficient.
  • Combination values including:
  • the combined value of the first dimensional attribute corresponding to each second dimensional attribute of the target user is obtained according to the product of the demand level value of the first dimensional attribute corresponding to the target user and the corresponding correlation coefficient.
  • the determining the total value of the second dimensional attribute of each skin care product according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product includes:
  • the second dimension attribute value corresponding to the target user is taken as the corresponding second dimension attribute value in each skin care product, and the sum of all the second dimension attribute values in each skin care product is calculated to obtain the first dimension of each skin care product.
  • the product information set in the product information set is determined according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product in the preset product information set Before the total value of the second dimension attribute of each skin care product, the method further includes:
  • the method further includes:
  • a skin care product recommendation device comprising:
  • the detection module is configured to obtain an image containing the skin of the target user, and process the image to obtain at least one first-dimensional attribute value corresponding to the target user's skin, wherein each first-dimensional attribute is used to represent one Skin indicator attributes;
  • the first processing module is configured to obtain at least one second dimensional attribute value corresponding to the target user according to the first dimensional attribute value and the correlation relationship between the first dimensional attribute and the second dimensional attribute, wherein:
  • Each of the second-dimensional attributes is used to represent a product index attribute related to the skin;
  • the second processing module is configured to determine each skin care product in the product information set according to at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product in the preset product information set The total value of the second dimension attribute of the product;
  • the target determination module is configured to determine the skin care product to be pushed corresponding to the target user according to the total value of the second dimension attribute of each skin care product in the product information set and preset pushing conditions.
  • a computer device includes a memory and a processor, the memory stores a computer program, and the processor executes the steps of the method when the computer program is executed.
  • a computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method are realized.
  • the above-mentioned product information processing method, device, computer equipment, and storage medium process the acquired image containing the target user’s skin to obtain the first dimension attribute value corresponding to the target user’s skin, and then according to the first dimension attribute and the second dimension attribute.
  • the correlation relationship between the target users is obtained, and the second dimension attribute value corresponding to the target user is obtained, and the total value of the second dimension attribute of each skin care product is obtained according to the second dimension attribute value and the second dimension attribute situation of each skin care product.
  • the total value of the second-dimensional attributes of the product combined with the push conditions can filter out the skin care products to be pushed corresponding to the target user.
  • the attribute value of the target user is skin combined with the correlation between the skin index attributes and the product index attributes, the first skin care product is evaluated.
  • the two-dimensional attribute value makes the quantification of the attributes of skin care products more accurate, and the filtered skin care product information to be pushed is more accurate and more in line with the needs of target users.
  • Figure 1 is an application environment diagram of a product information processing method in an embodiment
  • Figure 2 is a flowchart of a product information processing method in an embodiment
  • Fig. 3 is a flowchart of a method of generating a skin attribute evaluation model in an embodiment
  • FIG. 4 is a flowchart of a method of generating a skin attribute evaluation model in another embodiment
  • Figure 5 is a flowchart of a product information processing method in another embodiment
  • Figure 6 is a structural block diagram of a product information processing device in an embodiment
  • Fig. 7 is a block diagram of the internal structure of a computer device in an embodiment.
  • the product information processing method provided by the embodiment of the present application can be applied to the application environment as shown in FIG. 1.
  • the computer device 102 can process the image containing the user's skin to obtain the skin index attribute value of the user's skin.
  • the product index attribute value corresponding to the user skin can be obtained, according to
  • the product index attribute value corresponding to the user’s skin and the product index attribute of each skin care product in the preset product information set are calculated to obtain the total value of the product index attribute of each skin care product in the product information set.
  • the total value of the product index attributes of each skin care product combined with preset pushing conditions can determine the skin care product to be pushed corresponding to the user.
  • the skin care product to be pushed is determined according to the real skin index attribute value of the user's skin.
  • the pushed skin care product is more in line with the actual needs of the user's skin, and the accuracy of information push is improved.
  • the computer device 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and may also be an independent server or a server cluster composed of multiple servers.
  • Fig. 2 is a flowchart of a product information processing method in an embodiment. As shown in Fig. 2, a method for processing product information, taking the computer device running in Fig. 1 as an example, the method includes:
  • Step 202 Obtain an image containing the skin of the target user, and process the image to obtain at least one first-dimensional attribute value corresponding to the target user's skin, wherein each first-dimensional attribute is used to represent a skin indicator attribute.
  • the target user refers to the user who needs to push skin care product information.
  • the target user skin may be the target user's facial skin, body skin, etc.
  • Body skin may include arm skin, hand skin, abdomen, back, leg skin, etc., and is not limited thereto.
  • the target user can be captured by a camera on a computer device or an independent camera or a camera on other equipment to obtain an image containing the skin of the target user.
  • the image containing the skin of the target user may be an image taken in real time, or an image taken within a preset time period from the recommended time of use by the user.
  • the image may be at least one of a facial image, a hand image, an abdominal image, and the like.
  • the image can be processed by the trained skin attribute evaluation model to obtain at least one first dimensional attribute value corresponding to the target user's skin.
  • the first dimension attribute is used to characterize a skin index attribute.
  • the skin index attribute refers to the attribute used to represent the skin condition.
  • the skin index attributes may include, but are not limited to: acne, wrinkles, pores, dry oiliness, stains, blackheads, dark circles, etc.
  • skin index attributes may include but are not limited to: acne, cracked skin, pores, dry oiliness, stains, etc.
  • the first dimension attribute value refers to the score obtained by quantifying the state of the skin index attribute. For example, taking the skin index attribute of acne as an example, when there is no acne, the value of the attribute value of acne is 10 points, and the number of acne is less than the first preset value, then the skin index of acne The attribute value is 8 points, and the number of acne is greater than the first preset value and less than the second preset value, then the attribute value of the skin index of acne is 6 points, where the first preset value is less than the second preset value , And so on.
  • the corresponding scores can also be configured according to the size of the acne area.
  • the skin professional evaluators can also mark the scores of the skin index attributes in the existing images, and use them as samples to train to obtain the skin attribute evaluation model.
  • the evaluation model directly recognizes and obtains the attribute value of the skin index in the image.
  • Step 204 According to the first dimension attribute value and the correlation relationship between the first dimension attribute and the second dimension attribute, at least one second dimension attribute value corresponding to the target user is obtained, wherein each of the second dimension attributes Attributes are used to characterize a product index attribute related to the skin.
  • the correlation relationship between the first dimension attribute and the second dimension attribute refers to the correlation relationship between each first dimension attribute and each second dimension attribute.
  • the correlation coefficient that characterizes the correlation between each first dimensional attribute and each second dimensional attribute can be pre-configured.
  • the correlation coefficient between the first dimension attribute and the second dimension attribute is used to quantify the relationship between the two to facilitate subsequent calculations.
  • the second dimension attribute is used to characterize the product index attributes related to the skin.
  • Product indicator attributes are attributes used to represent product characteristics (such as ingredients and/or efficacy, etc.).
  • Product index attributes can include: additives, whether it is suitable for acne skin, whether it has a secondary cleaning effect, preservative content, whether it is suitable for allergic skin, type of ingredients, efficacy, decontamination, whether it contains oil, and whether there are hidden dangers of acne and allergy , Moisturizing, UVA (Ultraviolet A) ability, and UVB (ultraviolet radiation b, outdoor ultraviolet) ability.
  • the UVA band is a part of the ultraviolet wavelength division, with a wavelength of 320 to 420 nm.
  • Additives such as: no synthetic fat, no alcohol, no coloring, coloring, no fragrance, more fragrance, less fragrance, certain fragrance, no vegetable oil, etc.
  • acne skin quality can be used, acne skin quality used with caution, acne skin quality selection, etc.
  • Preservative content such as: more preservatives, less preservatives, certain preservatives, halogen-containing preservatives, no preservatives, alternative preservatives, etc.
  • allergic skin can be used, allergic skin used with caution, allergic skin selection, etc.
  • Types of ingredients such as: rich in natural ingredients, mint ingredients, surfactants, polyols, formaldehyde releasers, yeast ingredients, waxes, parabens, ceramides, liposomes, etc. .
  • Efficacy such as: anti-acne, moisturizing, anti-glycation, anti-inflammatory, anti-oxidation, anti-wrinkle, oil control, keratin conditioning, healing, repair, whitening, tanning, detoxification, soothing, moisturizing degree, strong moisturizing, moisturizing General, clean, strong cleaning power, weak cleaning power, average cleaning power, etc.
  • Decontamination for example: strong decontamination power, average decontamination power, etc.
  • Does it contain grease for example: no silicone oil, no mineral oil, medium synthetic grease, medium mineral oil, medium vegetable oil, small amount of synthetic grease, small amount of mineral oil, small amount of vegetable oil, no wax, etc.
  • Moisturizing degree such as: strong moisturizing degree, weak moisturizing degree, average absorption, etc.
  • Anti-UVA ability such as: PA+, PA++, PA+++, etc.
  • Anti-UVB ability such as: SPF10+, SPF20+, SPF30+, etc.
  • the product index attributes may include but are not limited to: no synthetic fats, ceramides, available for acne-prone skin, use with caution for acne-prone skin, moisturizing, strong cleaning ability, weak cleaning ability, anti-oxidation, whitening, etc. .
  • the attribute value of the second dimension may be a score obtained by quantifying the product index attribute.
  • the computer device can obtain the product of each first dimension attribute value corresponding to the second dimension attribute and the correlation coefficient between the first dimension attribute and the second dimension attribute to obtain each first dimension corresponding to the second dimension attribute.
  • the combined value of the attribute can be obtained by summing the combined value of each first dimensional attribute corresponding to the second dimensional attribute to obtain the second dimensional attribute value.
  • Each second-dimensional attribute value can be obtained by summing the combined values of each first-dimensional attribute corresponding to the second dimension attribute.
  • the second dimension attribute value may also be weighted and summed according to the combined value of each first dimension attribute to obtain the second dimension attribute value.
  • Step 206 Determine the total value of the second dimension attribute of each skin care product in the product information set according to the at least one second dimension attribute value corresponding to the target user and the preset second dimension attribute of each skin care product in the product information set .
  • the product information of at least one skin care product is recorded in the product information collection.
  • the second-dimensional attributes of skin care products refer to the second-dimensional attributes contained in skin care products.
  • the total value of the second dimension attribute of each skin care product can be obtained by summing or weighting the attribute values of the second dimension included in the skin care product.
  • Step 208 Determine the skin care product to be pushed corresponding to the target user according to the total value of the second dimension attribute of each skin care product in the product information set and the preset pushing conditions.
  • the preset push conditions can be configured as needed.
  • the push conditions can be based on the total value. For example, if the higher the score, the better, the push conditions can include any of the following: push the skin care product with the highest total value; push the preset number of skin care products from high to low Products; push skin care products whose total value exceeds a preset threshold, etc. If the score is as low as possible, the push conditions can include any of the following: push skin care products with the lowest total value; push a preset number of skin care products with a total value from low to high; push skin care products with a total value below the preset threshold Pin etc.
  • the non-correlation in the correlation relationship is considered, specifically: when the total value of multiple skin care products is the same , Acquire the number of unrelated relationships in the correlation relationship between the second dimension attribute and the first dimension attribute of multiple products with the same total value, and select the product with a large number of unrelated relationships as the product to be pushed.
  • the product information processing method in this embodiment processes the acquired image containing the target user skin to obtain the first dimension attribute value corresponding to the target user skin, and then according to the correlation between the first dimension attribute and the second dimension attribute Relationship, the second dimension attribute value corresponding to the target user is obtained, and the total value of the second dimension attribute of each skin care product is obtained according to the second dimension attribute value and the second dimension attribute situation of each skin care product.
  • the total value of the dimension attribute combined with the push conditions can filter out the skin care products to be pushed corresponding to the target user, and the second dimension attribute value of the skin care product is evaluated according to the attribute value of the target user’s skin and the correlation between the skin index attribute and the product index attribute.
  • the quantification of the scores of skin care products is more accurate, and the selected skin care products to be pushed are more accurate and more in line with the needs of target users.
  • the obtaining an image containing the skin of the target user, and processing the image to obtain at least one first dimension attribute value corresponding to the skin of the target user includes: obtaining an image containing the skin of the target user, and inputting the image At least one first-dimensional attribute value corresponding to the target user's skin is obtained by the trained skin attribute evaluation model, and the skin attribute evaluation model is obtained by training based on an image containing the first-dimensional attribute value.
  • the skin attribute evaluation model is obtained by training based on the image containing the attribute value of the first dimension as a training sample.
  • the image containing the attribute value of the first dimension refers to first selecting the image containing the user's skin, and then marking the corresponding value of the skin indicator attribute in the image to obtain the image with the attribute value of the first dimension.
  • the skin attribute evaluation model may be a neural network model.
  • the neural network model may be a deep neural network model or a convolutional neural network model.
  • the first dimension attribute value obtained by evaluating the image through the skin attribute evaluation model is more accurate.
  • Fig. 3 is a flowchart of a method of generating a skin attribute evaluation model in an embodiment. As shown in Fig. 3, in one embodiment, the method for generating the skin attribute evaluation model includes:
  • Step 302 Input the sample image containing the attributes of the first dimension into the skin attribute evaluation model containing the initial values of the weighted parameters.
  • the skin attribute evaluation model may be a convolutional neural network model. First, the weight parameters of the convolutional neural network model are given initial values, and then the convolutional neural network model with the initial values is used to train the image with the first dimension attribute score value.
  • Step 304 Extract the skin index attribute in the sample image through the skin attribute evaluation model, and calculate the actual value of the skin index attribute.
  • the skin attribute evaluation model can first extract each local feature of the skin in the sample image, and then extract the skin index attribute from each local feature, and then use the skin attribute evaluation model to detect the skin index attribute in the sample image.
  • the actual value of the skin indicator attribute may be a global score value, and the global score value refers to a score value representing a certain skin index attribute in the entire image.
  • Step 306 Determine a loss function value according to the actual value and the reference value.
  • the reference value of each skin index attribute refers to the score value for marking each skin index attribute in the sample image.
  • the value of the loss function can be obtained by calculating the difference between the actual value and the reference value, or the value of a linear function containing the difference between the actual value and the reference value can be used as the value of the loss function.
  • Step 308 Adjust the initial value of the weight parameter of the skin attribute evaluation model according to the loss function value, until the loss function value meets a preset condition, obtain the weight parameter target value of the skin attribute evaluation model.
  • the loss function value is greater than the expected value
  • adjust the weight parameters of the skin attribute evaluation model for example, the network layer of the convolutional neural network adjusts the weight layer by layer
  • the preset condition such as less than the expected value
  • the skin attribute evaluation model is trained through pre-built sample images containing skin index attributes, and a skin attribute evaluation model that meets the requirements can be obtained, which facilitates subsequent accurate identification of the skin index attributes in the captured image. Detection value.
  • Fig. 4 is a flowchart of a method of generating a skin attribute evaluation model in an embodiment. As shown in Fig. 4, in one embodiment, the method of generating the skin attribute evaluation model includes:
  • Step 402 Input the sample image including the skin attribute index into the skin attribute evaluation model including the initial value of the weighted parameter.
  • the skin attribute evaluation model may be a convolutional neural network model. First, the weight parameters of the convolutional neural network model are given initial values, and then the convolutional neural network model with the initial values is used to train the image with the first dimension attribute score value.
  • Step 404 Extract the skin index attribute in the sample image through the skin attribute evaluation model.
  • the skin attribute evaluation model may first extract each local feature of the skin in the sample image, and then extract the skin index attribute from each local feature.
  • Step 406 Identify the sample area corresponding to the skin index attribute in the sample image.
  • the target detection algorithm can be used to identify the sample area corresponding to the skin index attribute in the template image.
  • step 404 and step 406 can be performed simultaneously.
  • Step 408 Calculate the actual value of the skin index attribute according to the skin index attribute and the corresponding sample area.
  • the score value of the skin index attribute of the sample area is detected as the actual score value of the skin index attribute.
  • the total score or average score or weighted score of the skin index attribute scores of the multiple sample areas corresponding to the skin index attribute is used as the actual skin index attribute value.
  • Step 410 Determine a loss function value according to the actual value and the reference value.
  • the reference score value of each skin index attribute refers to the score value of each skin index attribute marked in the sample image.
  • the value of the loss function can be obtained by calculating the difference between the actual score value and the reference score value, or the value of a linear function containing the difference between the actual score value and the reference score value as the loss function value.
  • Step 412 Adjust the initial value of the weight parameter of the skin attribute evaluation model according to the loss function value, until the loss function value meets a preset condition, obtain the weight parameter target value of the skin attribute evaluation model.
  • the loss function value is greater than the expected value
  • adjust the weight parameters of the skin attribute evaluation model for example, the network layer of the convolutional neural network adjusts the weight layer by layer
  • the preset condition such as less than the expected value
  • the skin attribute evaluation model is trained through pre-built sample images containing skin index attributes, and the skin index attributes are evaluated by identifying the sample area corresponding to the skin index attributes and the characteristics of the skin index attributes.
  • the actual score value is more accurate, and the loss function value obtained is more accurate, and the value of the weight parameter is adjusted more accurately.
  • the skin attribute evaluation model that meets the needs can be obtained quickly, which facilitates the subsequent accurate identification of the skin index attributes in the captured image The detection value.
  • the product information processing method After the total value of the second dimensional attribute of each skin care product is determined based on the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product, the product information processing method It also includes: normalizing the total value of the second dimension attribute of each skin care product.
  • the normalization process may select the highest value or the lowest value or some other value in the total value as the reference value, and then normalize the total value of each second dimension attribute according to the reference value. For example, if the highest total value is actually 80, it is defined as 100, and other points are adjusted accordingly; or the lowest total value is actually 30, defined as 60, and other points are adjusted correspondingly through normalization processing, which is conducive to parallel comparison The matching degree of skin care products recommended for different objects.
  • the normalization process for the total value of the second dimension attribute of each skin care product includes: selecting the highest value from the total value of the second dimension attribute of each skin care product in the product information set Or the lowest value is used as the reference value, and the total value of the second dimension attribute of each skin care product is normalized according to the reference value.
  • the calculation is convenient by selecting the highest value or the lowest value as the reference value.
  • the normalization process for the total value of the second dimension attribute of each skin care product includes: corresponding to the highest value of the total value of the second dimension attribute of each skin care product in the product information set To the upper limit of the preset range, the lowest value of the total value of the second dimension attribute of each skin care product in the product information set corresponds to the lower limit of the preset range, and then each item in the product information set
  • the total value between the highest value and the lowest value of the total value of the second dimension attribute of each skin care product corresponds to the preset range.
  • the highest total value is actually 98, which is defined as 100
  • the lowest total value is actually 44, which is defined as 60, and other points are adjusted accordingly.
  • the obtaining at least one second dimensional attribute value corresponding to the target user according to the first dimensional attribute value in combination with the correlation relationship between the first dimensional attribute and the second dimensional attribute includes: obtaining a characterization The correlation coefficient of the correlation relationship between the first dimension attribute and the second dimension attribute; according to the value of each first dimension attribute corresponding to the target user and the corresponding correlation coefficient, the corresponding correlation coefficient of each second dimension attribute of the target user is obtained The combined value of each first dimensional attribute of the target user; at least one second dimensional attribute value of the target user is obtained according to the combined value of each first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the correlation relationship includes: a positive correlation relationship, a negative correlation relationship, and a non-correlation relationship, wherein the positive correlation relationship is used to indicate that the second dimension attribute can enhance the first dimension attribute, The negative correlation is used to indicate that the second dimensional attribute can weaken the first dimensional attribute, and the no correlation is used to indicate that the second dimensional attribute can have no influence on the first dimensional attribute.
  • the positive correlation relationship can be configured as a multi-level positive correlation relationship as needed.
  • the multi-level positive correlation relationship includes the first-level positive correlation, the second-level positive correlation, and the third-level positive correlation.
  • Each level of positive correlation indicates that the second dimension attribute enhances the influence of the first dimension attribute, or according to The first level of positive correlation, the second level of positive correlation, the third level of positive correlation, etc. from low to high level, each level of positive correlation indicates that the second dimension attribute enhances the influence of the first dimension attribute weaker .
  • the negative correlation relationship can be configured as a multi-level negative correlation relationship according to needs.
  • the multi-level positive correlation relationship includes the first-level negative correlation, the second-level negative correlation, and the third-level negative correlation...; can follow the first-level negative correlation Relationship, second-level negative correlation, third-level negative correlation, etc. from low to high.
  • Each negative relationship indicates that the second dimension attribute weakens the stronger the influence of the first dimension attribute, or according to the first Level negative correlation, second level negative correlation, third level negative correlation, etc. from low to high level, each level of negative correlation indicates that the second dimension attribute weakens the weaker the influence of the first dimension attribute.
  • the correlation relationship may also be expressed in the form of characters or numbers, for example, it may be expressed in +++, ++, +, 0, -, --, ---, etc.
  • +++, ++, + indicate that the second dimension attribute enhances the first dimension attribute
  • the second dimension attribute represented by "+++” enhances the influence of the first dimension attribute stronger than the first dimension attribute represented by "++”
  • Two-dimensional attributes enhance the influence of the first-dimensional attributes.
  • the second dimension attribute represented by "++” enhances the influence of the first dimension attribute stronger than the second dimension attribute represented by "+” enhances the influence of the first dimension attribute.
  • the second dimension attribute represented by "---" weakens the influence of the first dimension attribute stronger than the second dimension attribute represented by "--” weakens the influence of the first dimension attribute.
  • the second dimension attribute represented by "--" weakens the influence of the first dimension attribute stronger than the second dimension attribute represented by "-” weakens the influence of the first dimension attribute.
  • the correlation relationship between the first dimension attribute and the second dimension attribute may be as shown in Table 1.
  • O means no correlation; ++ means strong positive correlation; + means positive correlation;-means negative correlation; - means strong negative correlation.
  • the correlation coefficient representing the correlation relationship between the first dimension attribute and the second dimension attribute can be configured as required. For example, no correlation, the correlation coefficient is 0; positive correlation, the correlation coefficient is 1; strong positive correlation, the correlation coefficient is 2; negative correlation, the correlation coefficient is -1; strong negative correlation, correlation The coefficient is -2 and so on.
  • the product value of each first dimension attribute value corresponding to the target user and the corresponding correlation coefficient may be used as the combined value of each first dimension attribute corresponding to the target user.
  • obtaining the combined value of each first dimensional attribute corresponding to each second dimensional attribute of the target user according to each first dimensional attribute value corresponding to the target user and the corresponding correlation coefficient includes : According to the attribute value of each first dimension corresponding to the target user, the weight coefficient of each first dimension attribute and the corresponding correlation coefficient, each first dimension attribute corresponding to each second dimension attribute of the target user is obtained The combined value of.
  • a corresponding weight coefficient can be configured for each first dimension attribute.
  • the first dimension attributes include acne, pores, wrinkles, dry oiliness, stains, blackheads and dark circles, and the weight coefficients of the configurations are shown in Table 2.
  • the weight coefficients in Table 2 can be adjusted as needed.
  • the configuration of the weight coefficient of the first dimension attribute makes the value of the correlation between the first dimension attribute and the second dimension attribute more reasonable.
  • each first dimension attribute value corresponding to the target user can be used as each first dimension attribute corresponding to each second dimension attribute of the target user
  • each first dimension attribute value corresponding to the target user, the weight coefficient of each first dimension attribute, and the corresponding correlation coefficient are obtained to obtain each second dimension attribute of the target user.
  • the combined value of the first dimension attribute includes: converting each first dimension attribute value corresponding to the target user into the ability level value of each first dimension attribute corresponding to the target user; according to the first dimension attribute value corresponding to the target user. The product of the ability level value of the dimensional attribute, the weight coefficient of the first dimensional attribute and the corresponding correlation coefficient obtains the combined value of the first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the corresponding relationship between the first dimension attribute value and the ability level value can be established in advance, and the first dimension attribute value can be converted into the ability level value according to the corresponding relationship.
  • the value can be from low to low Push high.
  • the ability level value reflects the ability of the user's skin.
  • Table 3 The corresponding relationship between the attribute value of the first dimension and the ability level value can be shown in Table 3.
  • each first dimension attribute value corresponding to the target user, the weight coefficient of each first dimension attribute, and the corresponding correlation coefficient are obtained to obtain each second dimension attribute of the target user.
  • the combined value of the first dimensional attribute includes: converting each first dimensional attribute value corresponding to the target user into the demand level value of each first dimensional attribute corresponding to the target user; The product of the demand level value of the dimensional attribute, the weight coefficient of the first dimensional attribute and the corresponding correlation coefficient obtains the combined value of the first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the conversion relationship between the first dimension attribute value and the demand level value can be configured as required.
  • the attribute value of the first dimension reflects the specific ability of the target user, the demand is opposite to the ability.
  • the attribute value of the first dimension can be converted into a demand level value.
  • the value can be pushed from high to low. .
  • the conversion relationship between the first dimension attribute value and the demand level value is shown in Table 4.
  • obtaining the combined value of each first dimensional attribute corresponding to each second dimensional attribute of the target user according to each first dimensional attribute value corresponding to the target user and the corresponding correlation coefficient includes: Convert each first dimensional attribute value corresponding to the target user into the ability level value of each first dimensional attribute corresponding to the target user; according to the ability level value and corresponding correlation coefficient of the first dimensional attribute corresponding to the target user The product of to obtain the combined value of the first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the corresponding relationship between the first dimension attribute value and the ability level value can be established in advance, and the first dimension attribute value can be converted into the ability level value according to the corresponding relationship.
  • the value can be from low to low Push high.
  • the ability level value reflects the ability of the user's skin.
  • Table 3 The corresponding relationship between the attribute value of the first dimension and the ability level value can be shown in Table 3.
  • obtaining the combined value of each first dimensional attribute corresponding to each second dimensional attribute of the target user according to each first dimensional attribute value corresponding to the target user and the corresponding correlation coefficient includes : Convert each first dimension attribute value corresponding to the target user into the demand level value of each first dimension attribute corresponding to the target user; according to the demand level value and corresponding correlation of the first dimension attribute corresponding to the target user The product of the coefficients obtains the combined value of the first dimension attribute corresponding to each second dimension attribute corresponding to the target user.
  • the conversion relationship between the first dimension attribute value and the demand level value can be configured as required.
  • the attribute value of the first dimension reflects the specific ability of the target user, the demand is opposite to the ability.
  • the attribute value of the first dimension can be converted into a demand level value.
  • the value can be pushed from high to low. .
  • the conversion relationship between the first dimension attribute value and the demand level value is shown in Table 4.
  • the demand level value can be positively correlated with the attribute value of the first dimension, that is, the larger the attribute value of the first dimension, the greater the demand level value; the demand level value can be negatively correlated with the attribute value of the second dimension. In other words, the larger the attribute value of the first dimension, the smaller the demand level value.
  • determining the total value of the second dimensional attribute of each skin care product according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product includes: corresponding to the target user
  • the second dimension attribute value of is used as the corresponding second dimension attribute value of each skin care product, and the sum of all the second dimension attribute values of each skin care product is calculated to obtain the total value of the second dimension attribute of each skin care product.
  • the sum of all the second-dimensional attribute values in a skin care product is added to obtain the total second-dimensional attribute value of the skin care product, and the calculation is simple.
  • the skin index attribute value of the user's skin and the corresponding weight coefficient may be formed into a skin index matrix of the user's skin. Construct a product index matrix for the second dimension attribute of each skin care product, that is, the product index attribute, encode it in one-hot mode, and then calculate the inner product of the two matrices to obtain the total value of the second dimension attribute of the corresponding skin care product.
  • the above-mentioned product information processing method further includes: obtaining product screening information, and determining the to-be-push corresponding to the target user according to the product screening information, the total value of the second dimension attribute of each skin care product, and preset push conditions Skin care products.
  • the product screening information may include product category, brand, origin, price range, push usage season (for example: spring, summer, autumn or winter, etc.)/use time (for example, morning, day, night, before bedtime, etc.).
  • the skin care products to be pushed can be screened out that better meet the needs of users.
  • Figure 5 is a flowchart of a product information processing method in another embodiment.
  • the difference between the product information processing method and the product information processing method in Figure 2 is that the candidate skin care products are first selected according to the product screening information, and then the total value of the candidate skin care products is calculated , Reducing the calculation of all skin care products in the product information collection, reducing the amount of calculation.
  • Step 502 Obtain an image containing the skin of the target user, and process the image to obtain at least one first-dimensional attribute value corresponding to the target user's skin, where each first-dimensional attribute is used to represent a skin indicator attribute.
  • Step 504 According to the first dimension attribute value and the correlation relationship between the first dimension attribute and the second dimension attribute, at least one second dimension attribute value corresponding to the target user is obtained, wherein each of the second dimension attributes Attributes are used to characterize a product index attribute related to the skin.
  • Step 506 Obtain product screening information, and filter candidate skin care products from the product information collection according to the product screening information.
  • the product information set is used to store skin care product data, which includes but is not limited to the second-dimensional attributes, category, brand, price, place of production, recommended season/time of use, and applicable people of each skin care product.
  • the product information collection may be a skin care product database.
  • the product screening information includes, but is not limited to, the second-dimensional attributes, category, brand, origin, price range, recommended season/time of use, and applicable people of the product.
  • the product screening information can be input by the target user, such as the product category, brand, origin, price range, use season/use time and other information entered by the user.
  • the product screening information may also be obtained by analyzing the user's historical usage data, and the historical usage data may be the type, brand, origin, price range, etc. of the skin care products used in the history of the user.
  • the product screening information may also be obtained by analyzing the user's personal attribute information.
  • Personal attribute information can be information such as identity and occupation. For example, according to the analysis of identity and occupation, the price range that can be consumed, favorite brands, and origins are obtained.
  • Step 508 Determine the second dimensional attribute of each skin care product in the candidate skin care products according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute value of each skin care product in the candidate skin care products Total value.
  • Step 510 Determine the skin care product to be pushed corresponding to the target user according to the total value of the second dimension attribute of each skin care product in the candidate skin care product and the preset pushing conditions.
  • step 506 may precede step 502 or step 504.
  • the product information processing method in this embodiment processes the acquired image containing the skin of the target user to obtain the first dimension attribute value corresponding to the target user skin, and then according to the correlation between the first dimension attribute and the second dimension attribute , Get the second dimension attribute value corresponding to the target user, filter out candidate skin care products according to the product screening information, and then obtain the candidate skin care products according to the second dimension attribute value and the second dimension attribute of each skin care product in the candidate skin care products The total value of the second dimension attribute of each skin care product.
  • the skin care product to be pushed corresponding to the target user can be filtered out, and the skin index attribute and product can be combined according to the attribute value of the target user’s skin
  • the correlation between the index attributes is evaluated to obtain the second-dimensional attribute value of the skin care product, and the score of the skin care product is more accurately quantified.
  • the selected skin care product to be pushed is more accurate and more in line with the needs of the target user.
  • the product information processing method further includes: displaying all or part of the attribute information of the skin care product to be pushed corresponding to the target user.
  • all the attribute information may include product brand, category, origin, price, ingredient type, efficacy, moisturizing degree, decontamination ability, additive status, usage method, and so on.
  • Part of the attribute information refers to the attribute information filtered as needed from all the attribute information.
  • the first dimension attribute value of A is shown in Table 5:
  • the ability level value and the demand level value refer to the previous example and are calculated according to the conversion relationship.
  • the basic value the demand level value * the weight coefficient; the results are shown in Table 6:
  • the correlation table of the first dimension attribute and the second dimension attribute is shown in Table 7 (the first rule table).
  • O means no correlation, with a score of 0; +, means a positive correlation, with a score of 1, ++, with a score of 2; -, means negative correlation, with a score of -1; --, the score Is -2.
  • the combined value of the first dimensional attribute corresponding to the second dimensional attribute of A the basic value of the first dimensional attribute of A * the corresponding correlation coefficient.
  • Price range 0-100 corresponds to A, 100-200 corresponds to B, 200-500 corresponds to C, 500-1000 corresponds to D, NA means unknown; 0 means no such attribute, + means there is this attribute.
  • the total value of the second dimension attribute of the skin care product the sum of all the second dimension attributes of the skin care product in the second dimension attribute values of the A, the calculation result is shown in Table 10.
  • the highest score push "Shiseido New White Brightening Moisturizing Gel”.
  • the push may also include other dimensions, such as having a specific second characteristic, product category, price range, origin, season/time of push use, and so on.
  • the screening of other dimensions can be carried out after the final result is obtained, or the product information collection can be screened at the beginning of the calculation to reduce the amount of calculation and improve efficiency.
  • Fig. 6 is a structural block diagram of a product information processing device in an embodiment.
  • a product information processing device includes a detection module 610, a first processing module 620, a second processing module 630, and a target determination module 640.
  • the detection module 610 is configured to obtain an image containing the skin of the target user, and process the image to obtain at least one first-dimensional attribute value corresponding to the target user's skin, wherein each first-dimensional attribute is used to represent a skin indicator attribute .
  • the first processing module 620 is configured to obtain at least one second dimensional attribute value corresponding to the target user according to the first dimensional attribute value in combination with the correlation relationship between the first dimensional attribute and the second dimensional attribute, where each The second dimension attribute is used to represent a product index attribute related to the skin.
  • the second processing module 630 is configured to determine the value of each skin care product in the product information set according to at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product in the preset product information set. The total value of the second dimension attribute.
  • the target determination module 640 is configured to determine the skin care product to be pushed corresponding to the target user according to the total value of the second dimension attribute of each skin care product in the product information set and the preset pushing conditions.
  • the product information processing device in this embodiment processes the acquired image containing the target user’s skin to obtain the first dimension attribute value corresponding to the target user’s skin, and then according to the correlation between the first dimension attribute and the second dimension attribute Relationship, obtain the second-dimensional attribute value corresponding to the target user, and obtain the total second-dimensional attribute of each skin care product according to the second-dimensional attribute value and the second-dimensional attribute situation of each skin care product in the preset product information set
  • the skin care product to be pushed corresponding to the target user can be filtered out, and the correlation relationship between the skin index attribute and the product index attribute can be evaluated according to the target user’s skin attribute value
  • the second-dimensional attribute value of skin care products can quantify the attributes of skin care products more accurately, and the selected skin care products to be pushed are more accurate and more in line with the needs of target users.
  • the detection module 610 is further configured to obtain an image containing the skin of the target user, and input the image into the trained skin attribute evaluation model to obtain at least one first dimensional attribute value corresponding to the target user's skin.
  • the skin attribute evaluation model is obtained by training based on an image containing the attribute value of the first dimension.
  • the device for processing information about pushing products further includes a training module.
  • the training module is used to input the sample image containing the first dimension attribute into the skin attribute evaluation model containing the initial value of the weighting parameter; extract the skin index attribute in the sample image through the skin attribute evaluation model, and calculate The actual value of the skin index attribute; determine the loss function value according to the actual value and the reference value; adjust the initial value of the weight parameter of the skin attribute evaluation model according to the loss function value until the loss function value meets a preset Under the condition, the target value of the weight parameter of the skin attribute evaluation model is obtained.
  • the training module is further used to input a sample image containing skin attribute indicators into a skin attribute evaluation model that includes initial values of weighted parameters; extract the skin in the sample image through the skin attribute evaluation model Index attribute; identify the sample area corresponding to the skin index attribute in the sample image; calculate the actual value of the skin index attribute according to the skin index attribute and the corresponding sample area; determine the loss according to the actual value and the reference value Function value; adjust the initial value of the weight parameter of the skin attribute evaluation model according to the loss function value, until the loss function value meets a preset condition, obtain the weight parameter target value of the skin attribute evaluation model.
  • the above product information processing device further includes a normalization processing module.
  • the normalization processing module is also used to determine the product information set according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product in the preset product information set After the total value of the second dimension attribute of each skin care product, the total value of the second dimension attribute of each skin care product is normalized.
  • the normalization processing module is further used to select the highest value or the lowest value from the total value of the second dimension attribute of each skin care product in the product information set as the reference value, and according to the reference value The total value of the second dimension attribute of each skin care product is normalized.
  • the normalization processing module is further used to correspond the highest value of the total value of the second dimension attribute of each skin care product in the product information set to the upper limit value of the preset range, and the The lowest value of the second dimension attribute total value of each skin care product in the product information set corresponds to the lower limit of the preset range, and then the second dimension attribute total value of each skin care product in the product information set The total value between the highest value and the lowest value of corresponds to the preset range.
  • the first processing module 620 includes a coefficient acquisition unit, a first calculation unit, and a second calculation unit.
  • the coefficient obtaining unit is used to obtain a correlation coefficient representing the correlation relationship between the first dimension attribute and the second dimension attribute.
  • the first calculation unit is configured to obtain the combined value of each first dimensional attribute corresponding to each second dimensional attribute of the target user according to each first dimensional attribute value corresponding to the target user and the corresponding correlation coefficient.
  • the second calculation unit is configured to obtain at least one second dimensional attribute value of the target user according to the combined value of each first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the first calculation unit is further configured to obtain each first dimension attribute value of the target user, the weight coefficient of each first dimension attribute, and the corresponding correlation coefficient. The combined value of each first dimension attribute corresponding to the second dimension attribute.
  • the first calculation unit is further configured to convert each first dimensional attribute value corresponding to the target user into a demand level value of each first dimensional attribute corresponding to the target user;
  • the product of the demand level value of the first dimensional attribute corresponding to the user, the weight coefficient of the first dimensional attribute, and the corresponding correlation coefficient obtains the combined value of the first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the first calculation unit is further configured to convert each first dimensional attribute value corresponding to the target user into a demand level value of each first dimensional attribute corresponding to the target user;
  • the product of the demand level value of the first dimensional attribute corresponding to the user and the corresponding correlation coefficient obtains the combined value of the first dimensional attribute corresponding to each second dimensional attribute of the target user.
  • the second processing module is further configured to use the second dimension attribute value corresponding to the target user as the corresponding second dimension attribute value in each skin care product, and obtain all the second dimension attributes in each skin care product.
  • the sum of the dimension attribute values is the total value of the second dimension attribute of each skin care product.
  • the above product information processing device further includes a screening information acquisition module and a screening module.
  • the screening information acquisition module is configured to determine each skin care product in the product information set according to the at least one second dimensional attribute value corresponding to the target user and the second dimensional attribute of each skin care product in the preset product information set. Before the total value of the second dimension attribute of the product, obtain the product screening information input by the target user;
  • the screening module is used to screen and obtain candidate skin care products from the product information collection according to the product screening information;
  • the second processing module is further configured to determine each skin care product in the candidate skin care product according to at least one second dimension attribute value corresponding to the target user and the second dimension attribute of each skin care product in the candidate skin care products The total value of the second dimension attribute.
  • the above-mentioned pushing product information processing device further includes a display module and a pushing module.
  • the display module is used to display all or part of the attribute information of the skin care product to be pushed corresponding to the target user.
  • the push module is used to push all or part of the attribute information of the skin care product to be pushed corresponding to the target user.
  • each module in the above product information processing device can be implemented in whole or in part by software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.
  • a computer device is provided.
  • the computer device may be a terminal or a server, and its internal structure diagram may be as shown in FIG. 7.
  • the computer equipment includes a processor, a memory, and a network interface connected through a system bus.
  • the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system and a computer program.
  • the internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer program is executed by the processor to realize a product information processing method.
  • the computer equipment may further include a display screen and an input device.
  • the display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen
  • the input device of the computer equipment can be a touch layer covered on the display screen, or it can be a button, a trackball or a touchpad set on the housing of the computer equipment , It can also be an external keyboard, touchpad, or mouse.
  • FIG. 7 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
  • the specific computer device may Including more or less parts than shown in the figure, or combining some parts, or having a different part arrangement.
  • a computer device includes a memory and a processor, the memory stores a computer program, and the processor executes the steps of a product information processing method when the computer program is executed.
  • the embodiment of the present application also provides a computer-readable storage medium.
  • the computer program can be stored in a non-volatile computer readable storage.
  • the medium when the computer program is executed, it may include the procedures of the above-mentioned method embodiments.
  • any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory.
  • the non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种产品信息处理方法包括：对包含目标用户皮肤的图像进行处理得到目标用户皮肤所对应的第一维度属性值；根据第一维度属性值，得到所述目标用户对应的第二维度属性值；根据目标用户对应的第二维度属性值确定每个护肤品的第二维度属性总值；根据每个护肤品的第二维度属性总值及推送条件确定目标用户对应的待推荐护肤品。

Description

产品信息处理方法、装置、计算机设备和存储介质
相关申请的交叉引用
本申请要求于2019年01月31日提交中国专利局、申请号为2019100990409、发明名称为“产品信息处理方法、装置、计算机设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机领域，特别是涉及一种产品信息处理方法、装置、计算机设备和存储介质。
背景技术
随着人们生活品质的提高，越来越多的人开始关注自身的皮肤状况。皮肤随着环境、饮食、护肤保养和人体个人因素等各种条件不同而呈现不同状态。目前人们选择护肤品通常是根据个人已往使用的情况来选择或其他使用者的推荐。个人已往使用的情况通常是根据个人喜好，选择自己经常使用的产品。其他使用者推荐通常是其他使用者根据自身某个产品的情况来推荐。两种方式所推荐的护肤品的准确性都较低。
发明内容
基于此，有必要针对上述技术问题，提供一种能够准确推送护肤品信息的产品信息处理方法、装置、计算机设备和存储介质。
一种产品信息处理方法，所述方法包括：
获取包含目标用户皮肤的图像，对所述图像进行处理得到所述目标用户皮肤所对应的至少一个第一维度属性值，其中，每个所述第一维度属性用于表征一个皮肤指标属性；
根据所述第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到所述目标用户对应的至少一个第二维度属性值，其中，每个所述第二维度属性用于表征一个与皮肤相关的产品指标属性；
根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值；
根据所述产品信息集合中的每个护肤品的第二维度属性总值及预设的推送条件确定所述目标用户对应的待推送护肤品。
在其中一个实施例中，所述获取包含目标用户皮肤的图像，对所述图像进行处理得到所述目标用户皮肤所对应的至少一个第一维度属性值，包括：
获取包含目标用户皮肤的图像，将所述图像输入到已训练的皮肤属性评估模型得到所述目标用户皮肤所对应的至少一个第一维度属性值，所述皮肤属性评估模型是根据包含有第一维度属性值的图像训练得到的。
在其中一个实施例中，所述皮肤属性评估模型的生成方式，包括：
将包含有第一维度属性的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；
通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性，并计算得到所述皮肤指标属性的实际值；
根据所述实际值与参考值确定损失函数值；
根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值。
在其中一个实施例中，所述皮肤属性评估模型的生成方式，包括：
将包含有皮肤属性指标的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；
通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性；
识别所述样本图像中皮肤指标属性所对应的样本区域；
根据所述皮肤指标属性及对应的样本区域计算得到所述皮肤指标属性的实际值；
根据所述实际值与参考值确定损失函数值；
根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值。
在其中一个实施例中，在所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值之后，所述方法还包括：
对所述产品信息集合中的每个护肤品的第二维度属性总值进行归一化处理。
在其中一个实施例中，所述对所述产品信息集合中的每个护肤品的第二维度属性总值进行归一化 处理，包括：
从所述产品信息集合中的每个护肤品的第二维度属性总值中选取最高值或最低值作为基准值，根据所述基准值对每个护肤品的第二维度属性总值进行归一化处理；或
将所述产品信息集合中的每个护肤品的第二维度属性总值的最高值对应到预设范围的上限值，将所述产品信息集合中的每个护肤品的第二维度属性总值的最低值对应到预设范围的下限值，再将位于所述产品信息集合中的每个护肤品的第二维度属性总值的最高值和最低值之间的总值对应到所述预设范围内。
在其中一个实施例中，所述根据所述第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到所述目标用户对应的至少一个第二维度属性值，包括：
获取表征第一维度属性和第二维度属性之间相关性关系的相关性系数；
根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值；
根据所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值得到所述目标用户的至少一个第二维度属性值。
在其中一个实施例中，所述根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：
根据所述目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值。
在其中一个实施例中，所述根据所述目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：
将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；
根据所述目标用户对应的第一维度属性的需求等级值、第一维度属性的权重系数和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
在其中一个实施例中，所述根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：
将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；
根据所述目标用户对应的第一维度属性的需求等级值和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
在其中一个实施例中，所述根据所述目标用户对应的至少一个第二维度属性值和每个护肤品的第二维度属性确定每个护肤品的第二维度属性总值，包括：
将所述目标用户对应的第二维度属性值作为每个护肤品中对应的第二维度属性值，求取每个护肤品中全部的第二维度属性值之和，得到每个护肤品的第二维度属性总值。
在其中一个实施例中，在所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值之前，所述方法还包括：
获取产品筛选信息；
根据所述产品筛选信息从产品信息集合中筛选得到候选护肤品；
所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值，包括：
根据所述目标用户对应的至少一个第二维度属性值和所述候选护肤品中的每个护肤品的第二维度属性确定所述候选护肤品中的每个护肤品的第二维度属性总值。
在其中一个实施例中，所述方法还包括：
显示或推送所述目标用户对应的待推送护肤品的全部或部分属性信息。
一种护肤品推荐装置，所述装置包括：
检测模块，用于获取包含目标用户皮肤的图像，对所述图像进行处理得到所述目标用户皮肤所对应的至少一个第一维度属性值，其中，每个所述第一维度属性用于表征一个皮肤指标属性；
第一处理模块，用于根据所述第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到所述目标用户对应的至少一个第二维度属性值，其中，每个所述第二维度属性用于表征一个与皮肤相关的产品指标属性；
第二处理模块，用于根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值；
目标确定模块，用于根据所述产品信息集合中的每个护肤品的第二维度属性总值及预设的推送条件确定所述目标用户对应的待推送护肤品。
一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时所述方法的步骤。
一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现所述的方法的步骤。
上述产品信息处理方法、装置、计算机设备和存储介质，通过将获取的包含目标用户皮肤的图像进行处理得到目标用户皮肤对应的第一维度属性值，再根据第一维度属性和第二维度属性之间的相关性关系，得到目标用户对应的第二维度属性值，并根据第二维度属性值和每个护肤品的第二维度属性情况得到每个护肤品的第二维度属性总值，根据护肤品的第二维度属性总值结合推送条件可筛选出目标用户对应的待推送护肤品，根据目标用户皮肤的属性值结合皮肤指标属性和产品指标属性之间的相关性关系评估得到护肤品的第二维度属性值，对护肤品的属性量化更加准确，筛选得到的待推送护肤品信息更加准确，也更符合目标用户的需求。
附图说明
图1为一个实施例中产品信息处理方法的应用环境图；
图2为一个实施例中产品信息处理方法的流程图；
图3为一个实施例中皮肤属性评估模型的生成方式的流程图；
图4为另一个实施例中皮肤属性评估模型的生成方式的流程图；
图5为另一个实施例中产品信息处理方法的流程图；
图6为一个实施例中产品信息处理装置的结构框图；
图7为一个实施例中计算机设备的内部结构框图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
需要说明的是，本发明的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区 别类似的对象，而不必用于描述特定的顺序或先后次序。
本申请实施例提供的产品信息处理方法，可以应用于如图1所示的应用环境中。其中，计算机设备102可对包含用户皮肤的图像进行处理得到用户皮肤的皮肤指标属性值，根据皮肤指标属性与产品指标属性之间的相关性关系，可以得到用户皮肤对应的产品指标属性值，根据用户皮肤对应的产品指标属性值和预设的产品信息集合中的每个护肤品的产品指标属性计算得到该产品信息集合中的每个护肤品的产品指标属性总值，根据该产品信息集合中的每个护肤品的产品指标属性总值结合预设的推送条件可以确定用户对应的待推送护肤品。根据用户皮肤的真实皮肤指标属性值来确定待推送护肤品，推送的护肤品更加符合用户皮肤的实际需求，提高了信息推送的准确性。计算机设备102可以但不限于是各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备，还可以是独立的服务器或者是多个服务器组成的服务器集群来实现。
图2为一个实施例中产品信息处理方法的流程图。如图2所示，一种产品信息处理方法，以运行于图1中的计算机设备为例，所述方法包括：
步骤202，获取包含目标用户皮肤的图像，对该图像进行处理得到该目标用户皮肤所对应的至少一个第一维度属性值，其中，每个该第一维度属性用于表征一个皮肤指标属性。
其中，目标用户是指需要推送护肤品信息的用户。目标用户皮肤可为目标用户的面部皮肤、身躯皮肤等。身躯皮肤可包括手臂皮肤、手部皮肤、腹部、背部、腿部皮肤等，不限于此。
可通过计算机设备上的摄像头或者独立的摄像头或其他设备上的摄像头拍摄目标用户得到包含目标用户皮肤的图像。该包含目标用户皮肤的图像可为实时拍摄的图像，也可为距离用户使用推荐时刻在预设时长内所拍摄的图像。
图像可为面部图像、手部图像、腹部图像等至少一种。可通过训练好的皮肤属性评估模型对该图像进行处理得到该目标用户皮肤所对应的至少一个第一维度属性值。第一维度属性是用于表征一个皮肤指标属性。皮肤指标属性是指用来代表皮肤状况的属性。针对面部，皮肤指标属性可包括但不限于：痘痘、皱纹、毛孔、干油性、色斑、黑头、黑眼圈等。针对手部，皮肤指标属性可包括但不限于：痘痘、裂皮、毛孔、干油性、色斑等。
第一维度属性值是指根据皮肤指标属性的状态进行量化得到的分值。例如以痘痘这项皮肤指标属性为例，当没有痘痘时，则痘痘这一项皮肤指标属性值为10分，痘痘数量小于第一预设值，则痘痘这一项皮肤指标属性值为8分，痘痘数量大于第一预设值且小于第二预设值，则痘痘这一项皮肤指标属 性值为6分，其中，第一预设值小于第二预设值，可依次类推。此外，也可根据痘痘所在面积大小来配置对应的分值等，也可由皮肤专业评估人员对已有的图像中皮肤指标属性标记分值，作为样本进行训练得到皮肤属性评估模型，通过皮肤属性评估模型直接识别得到图像中皮肤指标属性值。
步骤204，根据该第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到该目标用户对应的至少一个第二维度属性值，其中，每个该第二维度属性用于表征一个与皮肤相关的产品指标属性。
其中，第一维度属性和第二维度属性之间的相关性关系是指每个第一维度属性和每个第二维度属性之间的相关性关系。可预先配置表征每个第一维度属性和每个第二维度属性之间的相关性关系的相关性系数。通过第一维度属性和第二维度属性之间的相关性系数来量化两者之间的关系，方便后续计算。
第二维度属性是用于表征与皮肤相关的产品指标属性。产品指标属性是用来代表产品特性(如成分和/或功效等)的属性。产品指标属性可包括：添加剂情况、痘肌是否适用、是否有二次清洁作用、防腐剂含量、过敏肤质是否适用、成分类型、功效、去污、是否含油脂、有无致痘致敏隐患、滋润度、防UVA(Ultra violet A，紫外线A)能力、防UVB(ultraviolet radiation b，户外紫外线)能力。其中，UVA波段是紫外线波长划分的一部分，波长320～420nm。
添加剂情况，例如：不含合成脂、不含酒精、不含色素、含色素、不含香精、香精较多、香精少、含一定香精、不含植物油等。
痘肌是否适用，例如：痘肌肤质可用、痘肌肤质慎用、痘肌肤质选用等。
是否有二次清洁作用：有或无。
防腐剂含量，例如：防腐剂较多、防腐剂少、含一定防腐剂、含卤素防腐剂、无防腐剂、替代性防腐剂等。
过敏肤质是否适用，例如：过敏肤质可用、过敏肤质慎用、过敏肤质选用等。
成分类型，例如：富含天然成分、含薄荷成分、含表面活性剂、含多元醇、含甲醛释放体、含酵母成分、含蜡、含尼泊金酯、含神经酰胺、含脂质体等。
功效，例如：抗痤疮、保湿、抗糖基化、抗炎、抗氧化、抗皱、控油、角质调理、愈合、修复、美白、美黑、排毒、舒敏、保湿程度、保湿性强、保湿性一般、清洁、清洁力强、清洁力弱、清洁力一般等。
去污，例如：去污力强、去污力一般等。
是否含油脂，例如：无硅油、无矿物油、中量合成脂、中量矿物油、中量植物油、少量合成脂、少量矿物油、少量植物油、无蜡等。
有无致痘致敏隐患，例如：无致痘隐患、无致敏隐患等。
滋润度，例如：滋润度强、滋润度弱、吸收性一般等。
防UVA能力，例如：PA+、PA++、PA+++等。
防UVB能力，例如：SPF10+、SPF20+、SPF30+等。
在一个实施例中，产品指标属性可包括但不限于：不含合成脂、含神经酰胺、痘肌肤质可用、痘肌肤质慎用、保湿、清洁能力强、清洁能力弱、抗氧化、美白等。
第二维度属性值可为对产品指标属性进行量化得到的分值。计算机设备可以根据第二维度属性所对应的每个第一维度属性值以及第一维度属性和第二维度属性之间的相关性系数求取乘积得到第二维度属性所对应的每个第一维度属性的组合值，再根据该第二维度属性所对应的每个第一维度属性的组合值求和可得到该第二维度属性值。每个第二维度属性值均可以根据第二维度属性所对应的每个第一维度属性的组合值求和得到。在其他实施例中，第二维度属性值也可以根据每个第一维度属性的组合值加权求和得到第二维度属性值。
步骤206，根据该目标用户对应的至少一个第二维度属性值和预设的产品信息集合中每个护肤品的第二维度属性确定该产品信息集合中每个护肤品的第二维度属性总值。
其中，产品信息集合中记录了至少一种护肤品的产品信息。护肤品的第二维度属性是指护肤品所包含的第二维度属性。可以将护肤品所包含的第二维度属性值求和或加权求和等方式得到每个护肤品的第二维度属性总值。
步骤208，根据该产品信息集合中每个护肤品的第二维度属性总值及预设的推送条件确定该目标用户对应的待推送护肤品。
其中，预设的推送条件可根据需要配置。推送条件可为根据总值进行推送，例如，若分值越高越好，则推送条件可包括以下任意一种：推送总值最高的护肤品；推送总值从高到低预设数量的护肤品；推送总值超过预设阈值的护肤品等。若分值越低越好，则推送条件可包括以下任意一种：推送总值最低的护肤品；推送总值从低到高预设数量的护肤品；推送总值低于预设阈值的护肤品等。
另外，需要说明的是，在一个实施例中，如果出现多个护肤品的总值相同，则考虑相关性关系中 的无相关性，具体为：在多个护肤品的总值相同的情况下，分别获取总值相同的多个产品的第二维度属性与第一维度属性的相关性关系中的无相关关系的数量，选择无相关关系数量多的产品作为待推送的产品。
本实施例中的产品信息处理方法，通过将获取的包含目标用户皮肤的图像进行处理得到目标用户皮肤对应的第一维度属性值，再根据第一维度属性和第二维度属性之间的相关性关系，得到目标用户对应的第二维度属性值，并根据第二维度属性值和每个护肤品的第二维度属性情况得到每个护肤品的第二维度属性总值，根据护肤品的第二维度属性总值结合推送条件可筛选出目标用户对应的待推送护肤品，根据目标用户皮肤的属性值结合皮肤指标属性和产品指标属性之间的相关性关系评估得到护肤品的第二维度属性值，对护肤品的评分量化更加准确，筛选得到的待推送护肤品更加准确，也更符合目标用户的需求。
在一个实施例中，该获取包含目标用户皮肤的图像，对该图像进行处理得到该目标用户皮肤所对应的至少一个第一维度属性值，包括：获取包含目标用户皮肤的图像，将该图像输入到已训练的皮肤属性评估模型得到该目标用户皮肤所对应的至少一个第一维度属性值，该皮肤属性评估模型是根据包含有第一维度属性值的图像训练得到的。
其中，皮肤属性评估模型是根据包含有第一维度属性值的图像作为训练样本进行训练得到的。包含第一维度属性值的图像是指先选择包含有用户皮肤的图像，然后对图像中的皮肤指标属性标记相应的值，得到带有第一维度属性值的图像。皮肤属性评估模型可为神经网络模型。神经网络模型可为深度神经网络模型或卷积神经网络模型等。
通过皮肤属性评估模型对图像进行评估得到的第一维度属性值更加准确。
图3为一个实施例中皮肤属性评估模型的生成方式的流程图。如图3所示，在一个实施例中，该皮肤属性评估模型的生成方式，包括：
步骤302，将包含有第一维度属性的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中。
具体地，通过采集大量、不同用户的面部皮肤照片或其他部位皮肤照片，按照预设标准，对每张照片标注第一维度属性的分数值，将带有第一维度属性分数值的图像作为样本图像。第一维度属性是指皮肤指标属性。皮肤属性评估模型可为卷积神经网络模型。先对卷积神经网络模型的权重参数赋予初始值，然后利用具有初始值的卷积神经网络模型对带有第一维度属性分数值的图像进行训练。
步骤304，通过该皮肤属性评估模型提取该样本图像中的皮肤指标属性，并计算得到该皮肤指标属性的实际值。
具体地，皮肤属性评估模型可以先提取样本图像中的皮肤的各个局部特征，然后从各个局部特征中提取出皮肤指标属性，然后利用皮肤属性评估模型对该样本图像中的皮肤指标属性进行检测得到该皮肤指标属性的实际值。该实际值可为全局分数值，全局分数值是指表征整个图像中某个皮肤指标属性的分数值。
步骤306，根据该实际值与参考值确定损失函数值。
具体地，每个皮肤指标属性的参考值是指样本图像中标记每个皮肤指标属性的分数值。可以将实际值与参考值求差得到损失函数值，也可以包含有实际值与参考值之差的线性函数的值作为损失函数值等。
步骤308，根据该损失函数值调整该皮肤属性评估模型的权重参数初始值，直到该损失函数值满足预设条件时，得到该皮肤属性评估模型的权重参数目标值。
具体地，损失函数值大于期望值时，调整皮肤属性评估模型的权重参数(例如，卷积神经网络的网络层逐层调整权重)，然后继续训练，直到损失函数值满足预设条件(如小于期望值)时，此时得到皮肤属性评估模型的权重参数目标值。
本实施例中，通过预先构建的包含有皮肤指标属性的样本图像对皮肤属性评估模型进行训练，可以得到满足需求的皮肤属性评估模型，方便后续准确的识别出拍摄的图像中的皮肤指标属性的检测值。
图4为一个实施例中皮肤属性评估模型的生成方式的流程图。如图4所示，在一个实施例中，该皮肤属性评估模型的生成方式，包括：
步骤402，将包含有皮肤属性指标的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中。
具体地，通过采集大量、不同用户的面部皮肤照片或其他部位皮肤照片，按照预设标准，对每张照片标注第一维度属性的分数值，将带有第一维度属性分数值的图像作为样本图像。第一维度属性是指皮肤指标属性。皮肤属性评估模型可为卷积神经网络模型。先对卷积神经网络模型的权重参数赋予初始值，然后利用具有初始值的卷积神经网络模型对带有第一维度属性分数值的图像进行训练。
步骤404，通过该皮肤属性评估模型提取该样本图像中的皮肤指标属性。
具体地，皮肤属性评估模型可以先提取样本图像中的皮肤的各个局部特征，然后从各个局部特征 中提取出皮肤指标属性。
步骤406，识别该样本图像中皮肤指标属性所对应的样本区域。
具体地，可以通过目标检测算法识别出样板图像中皮肤指标属性对应的样本区域。
在步骤402之后可以同时执行步骤404和步骤406。
步骤408，根据该皮肤指标属性及对应的样本区域计算得到该皮肤指标属性的实际值。
具体地，样本图像中某个皮肤指标属性对应的样本区域有一块，则检测得到该样本区域的皮肤指标属性的分数值作为该皮肤指标属性的实际分数值。
当样本图像中某个皮肤指标属性对应的样本区域有多个，则将皮肤指标属性对应的多个样本区域的皮肤指标属性分数值的总分或平均分或加权分作为该皮肤指标属性的实际值。
步骤410，根据该实际值与参考值确定损失函数值。
具体地，每个皮肤指标属性的参考分数值是指样本图像中标记每个皮肤指标属性的分数值。可以将实际分数值与参考分数值求差得到损失函数值，也可以包含有实际分数值与参考分数值之差的线性函数的值作为损失函数值等。
步骤412，根据该损失函数值调整该皮肤属性评估模型的权重参数初始值，直到该损失函数值满足预设条件时，得到该皮肤属性评估模型的权重参数目标值。
具体地，损失函数值大于期望值时，调整皮肤属性评估模型的权重参数(例如，卷积神经网络的网络层逐层调整权重)，然后继续训练，直到损失函数值满足预设条件(如小于期望值)时，此时得到皮肤属性评估模型的权重参数目标值。
本实施例中，通过预先构建的包含有皮肤指标属性的样本图像对皮肤属性评估模型进行训练，通过识别皮肤指标属性对应的样本区域以及皮肤指标属性的特征，然后进行检测得到的皮肤指标属性的实际分数值更加准确，进而得到的损失函数值更加准确，调整权重参数的值更加准确，可以较快的得到满足需求的皮肤属性评估模型，方便后续准确的识别出拍摄的图像中的皮肤指标属性的检测值。
在一个实施例中，在该根据该目标用户对应的至少一个第二维度属性值和每个护肤品的第二维度属性确定每个护肤品的第二维度属性总值之后，该产品信息处理方法还包括：对该每个护肤品的第二维度属性总值进行归一化处理。
具体地，归一化处理可以选取总值中最高值或最低值或某个其他值作为基准值，然后将各个第二维度属性总值按照基准值进行归一化。例如，最高总值实际为80，将其定义为100，其他分值对应调 整；或最低总值实际为30，将其定义为60，其他分值对应调整通过归一化处理，有利于平行比较针对不同对象推荐的护肤品的匹配度。
在一个实施例中，该对该每个护肤品的第二维度属性总值进行归一化处理，包括：从该产品信息集合中的每个护肤品的第二维度属性总值中选取最高值或最低值作为基准值，根据该基准值对每个护肤品的第二维度属性总值进行归一化处理。本实施例中，通过选取最高值或最低值作为基准值，计算方便。
在一个实施例中，该对该每个护肤品的第二维度属性总值进行归一化处理，包括：将该产品信息集合中的每个护肤品的第二维度属性总值的最高值对应到预设范围的上限值，将该产品信息集合中的每个护肤品的第二维度属性总值的最低值对应到预设范围的下限值，再将位于该产品信息集合中的每个护肤品的第二维度属性总值的最高值和最低值之间的总值对应到该预设范围内。例如，最高总值实际为98，将其定义为100，最低总值实际为44，将其定义为60，其他分值对应调整。通过将最高总值对应到预设范围的上限值，最低总值对应到预设范围的下限值，再将其他值映射到预设范围，计算方便，且方便查看。
在一个实施例中，该根据该第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到该目标用户对应的至少一个第二维度属性值，包括：获取表征第一维度属性和第二维度属性之间相关性关系的相关性系数；根据该目标用户对应的每个第一维度属性值及对应相关性系数得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值；根据该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值得到该目标用户的至少一个第二维度属性值。
在一个实施例中，所述相关性关系包括：正相关关系、负相关关系和无相关关系，其中，所述正相关关系用于表示所述第二维度属性能增强所述第一维度属性，所述负相关关系用于表示所述第二维度属性能减弱所述第一维度属性，所述无相关关系用于表示所述第二维度属性能对所述第一维度属性无影响。
其中，正相关关系可根据需要配置为多级正相关关系，多级正相关关系包括第一级正相关关系、第二级正相关关系、第三级正相关关系……；可按照第一级正相关关系、第二级正相关关系、第三级正相关关系等从低到高的级别顺序，每一级正相关关系表示第二维度属性增强第一维度属性的影响力越强，或按照第一级正相关关系、第二级正相关关系、第三级正相关关系等从低到高的级别顺序，每一级正相关关系表示第二维度属性增强第一维度属性的影响力越弱。负相关关系可根据需要配置为多 级负相关关系，多级正相关关系包括第一级负相关关系、第二级负相关关系、第三级负相关关系……；可按照第一级负相关关系、第二级负相关关系、第三级负相关关系等从低到高的级别顺序，每一级负相关关系表示第二维度属性减弱第一维度属性的影响力越强，或按照第一级负相关关系、第二级负相关关系、第三级负相关关系等从低到高的级别顺序，每一级负相关关系表示第二维度属性减弱第一维度属性的影响力越弱。
需要说明的是，所述相关性关系还可以是采用字符或数字等形式表示相关性关系，例如可采用+++、++、+、0、-、--、---等表示。其中，+++、++、+表示第二维度属性增强第一维度属性，“+++”所表示的第二维度属性增强第一维度属性的影响力强于“++”表示的第二维度属性增强第一维度属性的影响力。“++”所表示的第二维度属性增强第一维度属性的影响力强于“+”表示的第二维度属性增强第一维度属性的影响力。“---”所表示的第二维度属性减弱第一维度属性的影响力强于“--”表示的第二维度属性减弱第一维度属性的影响力。“--”所表示的第二维度属性减弱第一维度属性的影响力强于“-”表示的第二维度属性减弱第一维度属性的影响力。
具体地，第一维度属性和第二维度属性之间相关性关系可如表1所示。
表1
Figure PCTCN2020071631-appb-000001
其中，O，表示无相关性；++，表示强正相关；+，表示正相关；-，表示负相关；--，表示强负相 关。
表征第一维度属性和第二维度属性之间相关性关系的相关性系数可根据需要配置。例如无相关，则相关性系数为0；正相关，则相关性系数为1；强正相关，则相关性系数为2；负相关，则相关性系数为-1；强负相关，则相关性系数为-2等。第一维度属性和第二维度属性之间的相关性关系。
在一个实施例中，可以将目标用户对应的每个第一维度属性值与对应的相关性系数的乘积值作为目标用户对应的每个第一维度属性的组合值。每个第二维度属性所对应的每个第一维度属性的组合值求和得到目标用户的每个第二维度属性值。例如，痘痘的值为2分，含神经酰胺与痘痘的相关性系数为1，则含神经酰胺对应的痘痘的组合值为2*1＝2分，痘肌肤质可用与痘痘的相关性系数为2，则痘肌肤质可用对应的痘痘的组合值为2*2＝4。
在一个实施例中，该根据该目标用户对应的每个第一维度属性值及对应相关性系数得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：根据该目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数，得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值。
具体地，首先可以为每个第一维度属性配置对应的权重系数。例如第一维度属性包括痘痘、毛孔、皱纹、干油性、色斑、黑头和黑眼圈，其配置的权重系数如表2所示。
表2
  痘痘 毛孔 皱纹 干油性 色斑 黑头 黑眼圈
权重系数 10 8 9 10 9 9 9
表2中的权重系数可根据需要调整。第一维度属性的权重系数配置使得第一维度属性和第二维度属性之间的相关度的值更加合理。
可以将该目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数乘积作为该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值。例如，痘痘的值为2分，含神经酰胺与痘痘的相关性系数为1，痘痘的权重系数为10，则含神经酰胺对应的痘痘的组合值,2*1*10＝20分，痘肌肤质可用与痘痘的相关性系数为2，则痘肌肤质可用对应的痘痘的组合值为2*2*10＝40。
在一个实施例中，该根据该目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括： 将该目标用户对应的每个第一维度属性值转换为该目标用户对应的每个第一维度属性的能力等级值；根据该目标用户对应的第一维度属性的能力等级值、第一维度属性的权重系数和对应相关性系数的乘积得到该目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
具体地，可预先建立第一维度属性值与能力等级值之间的对应关系，根据该对应关系将第一维度属性值转换为能力等级值，后续推送护肤品时，则可以按照值从低到高进行推送。能力等级值反映的是用户皮肤所具有的能力。第一维度属性值与能力等级值之间的对应关系可如表3所示。
表3
第一维度属性值 能力等级值
8-10 3
5-7 2
0-4 1
例如，痘痘的值为2分，转换为能力等级值为1，含神经酰胺与痘痘的相关性系数为1，痘痘的权重系数为10，则含神经酰胺对应的痘痘的组合值,1*1*10＝10分，痘肌肤质可用与痘痘的相关性系数为2，则痘肌肤质可用对应的痘痘的组合值为1*2*10＝20。
在一个实施例中，该根据该目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：将该目标用户对应的每个第一维度属性值转换为该目标用户对应的每个第一维度属性的需求等级值；根据该目标用户对应的第一维度属性的需求等级值、第一维度属性的权重系数和对应相关性系数的乘积得到该目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
具体地，第一维度属性值与需求等级值的转换关系可根据需要配置。当第一维度属性值反映的是目标用户的具体能力时，则需求与能力相反，可以将第一维度属性值转换为需求等级值，后续推送护肤品时，则可以按照值从高到低推送。第一维度属性值与需求等级值的转换关系如表4所示。
表4
第一维度属性值 能力等级值 需求等级值
8-10 3 1
5-7 2 2
0-4 1 3
在一个实施例中，根据该目标用户对应的每个第一维度属性值及对应相关性系数得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：将该目标用户对应的每个第一维度属性值转换为该目标用户对应的每个第一维度属性的能力等级值；根据该目标用户对应的第一维度属性的能力等级值和对应相关性系数的乘积得到该目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
具体地，可预先建立第一维度属性值与能力等级值之间的对应关系，根据该对应关系将第一维度属性值转换为能力等级值，后续推送护肤品时，则可以按照值从低到高进行推送。能力等级值反映的是用户皮肤所具有的能力。第一维度属性值与能力等级值之间的对应关系可如表3所示。
例如，痘痘的值为2分，转换为能力等级值为1，含神经酰胺与痘痘的相关性系数为1，则含神经酰胺对应的痘痘的组合值,1*1＝1分，痘肌肤质可用与痘痘的相关性系数为2，则痘肌肤质可用对应的痘痘的组合值为1*2＝2。
在一个实施例中，该根据该目标用户对应的每个第一维度属性值及对应相关性系数得到该目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：将该目标用户对应的每个第一维度属性值转换为该目标用户对应的每个第一维度属性的需求等级值；根据该目标用户对应的第一维度属性的需求等级值和对应相关性系数的乘积得到该目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
具体地，第一维度属性值与需求等级值的转换关系可根据需要配置。当第一维度属性值反映的是目标用户的具体能力时，则需求与能力相反，可以将第一维度属性值转换为需求等级值，后续推送护肤品时，则可以按照值从高到低推送。第一维度属性值与需求等级值的转换关系如表4所示。
需要说明的是，需求等级值可与第一维度属性值正相关，也就是说，第一维度属性值越大，需求等级值越大；需求等级值可与第二维度属性值负相关，也就是说，第一维度属性值越大，需求等级值越小。
在一个实施例中，该根据该目标用户对应的至少一个第二维度属性值和每个护肤品的第二维度属性确定每个护肤品的第二维度属性总值，包括：将该目标用户对应的第二维度属性值作为每个护肤品中对应的第二维度属性值，求取每个护肤品中全部的第二维度属性值之和，得到每个护肤品的第二维度属性总值。将一个护肤品中的全部的第二维度属性值相加求和，得到该护肤品的第二维度属性总值， 计算简单。
在一个实施例中，可将用户皮肤的皮肤指标属性值和对应的权重系数形成用户皮肤的皮肤指标矩阵。对于每个护肤品的第二维度属性即产品指标属性构建一个产品指标矩阵，用one-hot方式编码，然后计算两个矩阵的内积得到对应的护肤品的第二维度属性总值。
在一个实施例中，上述产品信息处理方法还包括：获取产品筛选信息，根据产品筛选信息、每个护肤品的第二维度属性总值及预设的推送条件确定所述目标用户对应的待推送护肤品。其中，产品筛选信息可包括产品类别、品牌、产地、价格区间、推送使用季节(例如：春、夏、秋或冬等)/使用时间(例如早晨、白天、晚上、睡前等)等。
通过产品筛选信息、护肤品的第二维度属性总值和推送条件可筛选出更满足用户需求的待推送护肤品。
图5为另一个实施例中产品信息处理方法的流程图。如图5所示，在一个实施例中，该产品信息处理方法与图2中的产品信息处理方法的区别在于先根据产品筛选信息筛选出候选护肤品，然后再求取候选护肤品的总值，减少了对产品信息集合中全部护肤品进行计算，减少了计算量。
步骤502，获取包含目标用户皮肤的图像，对该图像进行处理得到该目标用户皮肤所对应的至少一个第一维度属性值，其中，每个该第一维度属性用于表征一个皮肤指标属性。
步骤504，根据该第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到该目标用户对应的至少一个第二维度属性值，其中，每个该第二维度属性用于表征一个与皮肤相关的产品指标属性。
步骤506，获取产品筛选信息，根据所述产品筛选信息从产品信息集合产品信息集合中筛选得到候选护肤品。
其中，产品信息集合用于存储护肤品数据，所述护肤品数据包括但不限于各护肤品的第二维度属性、类别、品牌、价格、产地、推荐使用季节/使用时间、适用人群等。产品信息集合可为护肤品数据库。所述产品筛选信息包括但不限于产品的第二维度属性、类别、品牌、产地、价格区间、推荐使用季节/使用时间、适用人群等。产品筛选信息可为目标用户输入的，例如用户输入的产品类别、品牌、产地、价格区间、使用季节/使用时间等信息。
在一个实施例中，产品筛选信息也可以为根据用户的历史使用数据分析得到的，历史使用数据可为用户历史所使用的护肤品的类型、品牌、产地、价格区间等。
在一个实施例中，产品筛选信息也可以为根据用户的个人属性信息分析得到的。个人属性信息可为身份、职业等信息。例如根据身份和职业分析得到可消费的价格区间、喜欢的品牌、产地等。
步骤508，根据该目标用户对应的至少一个第二维度属性值和所述候选护肤品中的每个护肤品的第二维度属性确定所述候选护肤品中的每个护肤品的第二维度属性总值。
步骤510，根据该候选护肤品中的每个护肤品的第二维度属性总值及预设的推送条件确定该目标用户对应的待推送护肤品。
在其他实施例中，步骤506可在步骤502或步骤504之前。
本实施例中产品信息处理方法，通过将获取的包含目标用户皮肤的图像进行处理得到目标用户皮肤对应的第一维度属性值，再根据第一维度属性和第二维度属性之间的相关性关系，得到目标用户对应的第二维度属性值，根据产品筛选信息筛选出候选护肤品，然后根据第二维度属性值和候选护肤品中的每个护肤品的第二维度属性情况得到候选护肤品中每个护肤品的第二维度属性总值，根据护肤品的第二维度属性总值结合推送条件可筛选出目标用户对应的待推送护肤品，根据目标用户皮肤的属性值结合皮肤指标属性和产品指标属性之间的相关性关系评估得到护肤品的第二维度属性值，对护肤品的评分量化更加准确，筛选得到的待推送护肤品更加准确，也更符合目标用户的需求，且筛选出了候选护肤品，计算量小。
在一个实施例中，该产品信息处理方法还包括：显示该目标用户对应的待推送护肤品的全部或部分属性信息。
具体地，全部属性信息可包括产品品牌、类别、产地、价格、成分类型、功效、滋润度、去污能力、添加剂情况、使用方法等等。部分属性信息是指从全部属性信息中根据需要筛选的属性信息。通过显示待推送护肤品的全部属性信息可方便详细查看，显示部分属性信息可节省显示所需的空间等。
下面以用户甲为例进行说明。
甲的第一维度属性值如表5所示：
表5
  痘痘 毛孔 皱纹 干油性 色斑 黑头 黑眼圈
第一维度属性值 5 4 8 5 8 5 6
能力等级值和需求等级值参考前述示例，根据转换关系换算而得，基础值＝需求等级值*权重系数；结果如表6：
表6
  痘痘 毛孔 皱纹 干油性 色斑 黑头 黑眼圈
第一维度属性值 5 4 8 5 8 5 6
能力等级值 2 1 3 1 3 1 1
需求等级值 2 3 1 3 1 3 3
权重系数 10 8 9 10 9 9 9
基础值 20 24 9 30 9 27 27
表7
Figure PCTCN2020071631-appb-000002
第一维度属性和第二维度属性的相关性表如表7(第一规则表)。其中，O，表示无相关性，分值为0；+，表示正相关，分值为1，++，分值为2；-，表示负相关，分值为-1；--，分值为-2。
结合甲的第一维度属性的基础值，和上述第一规则表，得甲的各第二维度属性对应的第一维度属性的组合值。
甲的第二维度属性对应的第一维度属性的组合值＝甲的第一维度属性的基础值*相应的相关性系数。甲的各第二维度属性对应的第一维度属性总得分即各第二维度属性值＝SUM(甲的各第二维度属性对应的第一维度属性的组合值)。
计算后的结果如表8。
表8
  痘痘 毛孔 皱纹 干油性 色斑 黑头 黑眼圈 总得分
不含合成脂 0 0 0 0 0 0 0 0
含神经酰胺 20 24 0 10 9 0 0 63
痘肌肤质可用 40 0 0 0 0 0 0 40
痘肌肤质慎用 -40 0 0 0 0 0 0 -40
保湿 0 24 9 0 0 27 0 60
清洁能力强 20 48 0 0 0 54 0 122
清洁能力弱 -20 -48 0 0 0 -54 0 -122
抗氧化 20 0 18 0 0 0 27 65
美白 0 0 0 0 18 0 0 18
根据甲的各第二维度属性对应的第一维度属性总得分(即第二维度属性值)，结合护肤品数据的第二维度属性信息，计算各护肤品对应的第二维度属性总得分，如表9所示。
表9
Figure PCTCN2020071631-appb-000003
Figure PCTCN2020071631-appb-000004
价格区间：0-100对应A，100-200对应B，200-500对应C，500-1000对应D，NA表示未知；0表示无该属性，+表示有该属性。
护肤品的第二维度属性总值＝该护肤品具有的所有第二维度属性在甲的各第二维度属性值加和，计算结果如表10。
表10
Figure PCTCN2020071631-appb-000005
Figure PCTCN2020071631-appb-000006
Figure PCTCN2020071631-appb-000007
由上表10可知，各护肤品的总得分中，最高分为183，最低为40。
因此可根据之前预设的推送条件，例如：分数最高：推送“资生堂新透白美肌亮润保湿啫喱”。
可以理解的是，推送时除了分数这个维度外，还可包括其他维度，例如具有某一特定第二特征，产品类别，价格区间，产地，推送使用季节/时间等等。
其他维度的筛选可在获得最终结果后进行，也可以在获得开始计算之初，对产品信息集合进行筛选，以降低运算量，提高效率。
图6为一个实施例中产品信息处理装置的结构框图。如图6所示，一种产品信息处理装置，包括检测模块610、第一处理模块620、第二处理模块630和目标确定模块640。
检测模块610用于获取包含目标用户皮肤的图像，对该图像进行处理得到该目标用户皮肤所对应的至少一个第一维度属性值，其中，每个该第一维度属性用于表征一个皮肤指标属性。
第一处理模块620用于根据该第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到该目标用户对应的至少一个第二维度属性值，其中，每个该第二维度属性用于表征一个与皮肤相关的产品指标属性。
第二处理模块630用于根据该目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定该产品信息集合中的每个护肤品的第二维度属性总值。
目标确定模块640用于根据该产品信息集合中的每个护肤品的第二维度属性总值及预设的推送条件确定该目标用户对应的待推送护肤品。
本实施例中的产品信息处理装置，通过将获取的包含目标用户皮肤的图像进行处理得到目标用户皮肤对应的第一维度属性值，再根据第一维度属性和第二维度属性之间的相关性关系，得到目标用户对应的第二维度属性值，并根据第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性情况得到每个护肤品的第二维度属性总值，根据护肤品的第二维度属性总值结合推送条件可筛选出目标用户对应的待推送护肤品，根据目标用户皮肤的属性值结合皮肤指标属性和产品指标属性之间的相关性关系评估得到护肤品的第二维度属性值，对护肤品的属性量化更加准确，筛选得到的待推送护肤品更加准确，也更符合目标用户的需求。
在一个实施例中，检测模块610还用于获取包含目标用户皮肤的图像，将所述图像输入到已训练的皮肤属性评估模型得到所述目标用户皮肤所对应的至少一个第一维度属性值，所述皮肤属性评估模型是根据包含有第一维度属性值的图像训练得到的。
在一个实施例中，推送产品信息处理装置还包括训练模块。训练模块用于将包含有第一维度属性的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性，并计算得到所述皮肤指标属性的实际值；根据所述实际值与参考值确定损失函数值；根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值。
在一个实施例中，训练模块还用于将包含有皮肤属性指标的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性；识别所述样本图像中皮肤指标属性所对应的样本区域；根据所述皮肤指标属性及对应的样本区域计算得到所述皮肤指标属性的实际值；根据所述实际值与参考值确定损失函数值；根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值。
在一个实施例中，上述产品信息处理装置还包括归一化处理模块。该归一化处理模块还用于在所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定产品信息集合中的每个护肤品的第二维度属性总值之后，对所述每个护肤品的第二维度属性总值进行归一化处理。
在一个实施例中，该归一化处理模块还用于从所述产品信息集合中的每个护肤品的第二维度属性总值中选取最高值或最低值作为基准值，根据所述基准值对每个护肤品的第二维度属性总值进行归一化处理。
在一个实施例中，该归一化处理模块还用于将所述产品信息集合中的每个护肤品的第二维度属性总值的最高值对应到预设范围的上限值，将所述产品信息集合中的每个护肤品的第二维度属性总值的最低值对应到预设范围的下限值，再将位于所述产品信息集合中的每个护肤品的第二维度属性总值的最高值和最低值之间的总值对应到所述预设范围内。
在一个实施例中，第一处理模块620包括系数获取单元、第一计算单元和第二计算单元。
系数获取单元用于获取表征第一维度属性和第二维度属性之间相关性关系的相关性系数。
第一计算单元用于根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值。
第二计算单元用于根据所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值得到所述目标用户的至少一个第二维度属性值。
在一个实施例中，第一计算单元还用于根据所述目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值。
在一个实施例中，第一计算单元还用于将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；根据所述目标用户对应的第一维度属性的需求等级值、第一维度属性的权重系数和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
在一个实施例中，第一计算单元还用于将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；根据所述目标用户对应的第一维度属性的需求等级值和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
在一个实施例中，第二处理模块还用于将所述目标用户对应的第二维度属性值作为每个护肤品中对应的第二维度属性值，求取每个护肤品中全部的第二维度属性值之和，得到每个护肤品的第二维度属性总值。
在一个实施例中，上述产品信息处理装置还包括筛选信息获取模块和筛选模块。
筛选信息获取模块用于在所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定产品信息集合中的每个护肤品的第二维度属性总值之前，获取所述目标用户输入的产品筛选信息；
筛选模块用于根据所述产品筛选信息从产品信息集合中筛选得到候选护肤品；
第二处理模块还用于根据所述目标用户对应的至少一个第二维度属性值和所述候选护肤品中的每个护肤品的第二维度属性确定所述候选护肤品中的每个护肤品的第二维度属性总值。
在一个实施例中，上述推送产品信息处理装置还包括显示模块和推送模块。显示模块用于显示所述目标用户对应的待推送护肤品的全部或部分属性信息。推送模块用于推送目标用户对应的待推送护肤品的全部或部分属性信息。
关于产品信息处理装置的具体限定可以参见上文中对于产品信息处理方法的限定，在此不再赘述。上述产品信息处理装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是终端，也可以是服务器，其内部结构图可以如图7所示。该计算机设备包括通过系统总线连接的处理器、存储器和网络接口。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统和计算机程序。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种产品信息处理方法。在其他实施例中，该计算机设备还可包括显示屏和输入装置。该计算机设备的显示屏可以是液晶显示屏或者电子墨水显示屏，该计算机设备的输入装置可以是显示屏上覆盖的触摸层，也可以是计算机设备外壳上设置的按键、轨迹球或触控板，还可以是外接的键盘、触控板或鼠标等。
本领域技术人员可以理解，图7中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
本申请实施例还提供了一种计算机设备。一种计算机设备，包括存储器和处理器，该存储器存储有计算机程序，该处理器执行该计算机程序时产品信息处理方法的步骤。
本申请实施例还提供了一种计算机可读存储介质。一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时产品信息处理方法的步骤。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存 储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种产品信息处理方法，所述方法包括：
    获取包含目标用户皮肤的图像，对所述图像进行处理得到所述目标用户皮肤所对应的至少一个第一维度属性值，其中，每个所述第一维度属性用于表征一个皮肤指标属性；
    根据所述第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到所述目标用户对应的至少一个第二维度属性值，其中，每个所述第二维度属性用于表征一个与皮肤相关的产品指标属性；
    根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值；
    根据所述产品信息集合中的每个护肤品的第二维度属性总值及预设的推送条件确定所述目标用户对应的待推送护肤品。
  2. 根据权利要求1所述的方法，所述获取包含目标用户皮肤的图像，对所述图像进行处理得到所述目标用户皮肤所对应的至少一个第一维度属性值，包括：
    获取包含目标用户皮肤的图像，将所述图像输入到已训练的皮肤属性评估模型得到所述目标用户皮肤所对应的至少一个第一维度属性值，所述皮肤属性评估模型是根据包含有第一维度属性值的图像训练得到的。
  3. 根据权利要求2所述的方法，所述皮肤属性评估模型的生成方式，包括：
    将包含有第一维度属性的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；
    通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性，并计算得到所述皮肤指标属性的实际值；
    根据所述实际值与参考值确定损失函数值；
    根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值；或
    所述皮肤属性评估模型的生成方式，包括：
    将包含有皮肤属性指标的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；
    通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性；
    识别所述样本图像中皮肤指标属性所对应的样本区域；
    根据所述皮肤指标属性及对应的样本区域计算得到所述皮肤指标属性的实际值；
    根据所述实际值与参考值确定损失函数值；
    根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值。
  4. 根据权利要求1所述的方法，在所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值之后，所述方法还包括：
    对所述产品信息集合中的每个护肤品的第二维度属性总值进行归一化处理；
    所述对所述产品信息集合中的每个护肤品的第二维度属性总值进行归一化处理，可包括：
    从所述产品信息集合中的每个护肤品的第二维度属性总值中选取最高值或最低值作为基准值，根据所述基准值对每个护肤品的第二维度属性总值进行归一化处理；或
    将所述产品信息集合中的每个护肤品的第二维度属性总值的最高值对应到预设范围的上限值，将所述产品信息集合中的每个护肤品的第二维度属性总值的最低值对应到预设范围的下限值，再将位于所述产品信息集合中的每个护肤品的第二维度属性总值的最高值和最低值之间的总值对应到所述预设范围内。
  5. 根据权利要求1所述的方法，所述根据所述第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到所述目标用户对应的至少一个第二维度属性值，包括：
    获取表征第一维度属性和第二维度属性之间相关性关系的相关性系数；
    根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值；
    根据所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值得到所述目标用户的至少一个第二维度属性值。
  6. 根据权利要求5所述的方法，所述根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，包括：
    根据所述目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值；或
    将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求 等级值；
    根据所述目标用户对应的第一维度属性的需求等级值和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值；
    所述根据所述目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值，可包括：
    将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；
    根据所述目标用户对应的第一维度属性的需求等级值、第一维度属性的权重系数和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
  7. 根据权利要求1所述的方法，所述根据所述目标用户对应的至少一个第二维度属性值和每个护肤品的第二维度属性确定每个护肤品的第二维度属性总值，包括：
    将所述目标用户对应的第二维度属性值作为每个护肤品中对应的第二维度属性值，求取每个护肤品中全部的第二维度属性值之和，得到每个护肤品的第二维度属性总值。
  8. 根据权利要求1所述的方法，在所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值之前，所述方法还包括：
    获取产品筛选信息；
    根据所述产品筛选信息从产品信息集合中筛选得到候选护肤品；
    所述根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值，包括：
    根据所述目标用户对应的至少一个第二维度属性值和所述候选护肤品中的每个护肤品的第二维度属性确定所述候选护肤品中的每个护肤品的第二维度属性总值。
  9. 根据权利要求1所述的方法，所述方法还包括：
    显示或推送所述目标用户对应的待推送护肤品的全部或部分属性信息。
  10. 一种产品信息处理装置，所述装置包括：
    检测模块，设置为获取包含目标用户皮肤的图像，对所述图像进行处理得到所述目标用户皮肤所对应的至少一个第一维度属性值，其中，每个所述第一维度属性用于表征一个皮肤指标属性；
    第一处理模块，设置为根据所述第一维度属性值，结合第一维度属性和第二维度属性之间的相关性关系，得到所述目标用户对应的至少一个第二维度属性值，其中，每个所述第二维度属性用于表征一个与皮肤相关的产品指标属性；
    第二处理模块，设置为根据所述目标用户对应的至少一个第二维度属性值和预设的产品信息集合中的每个护肤品的第二维度属性确定所述产品信息集合中的每个护肤品的第二维度属性总值；
    目标确定模块，设置为根据所述产品信息集合中的每个护肤品的第二维度属性总值及预设的推送条件确定所述目标用户对应的待推送护肤品。
  11. 根据权利要求10所述的装置，所述检测模块还被设置为获取包含目标用户皮肤的图像，将所述图像输入到已训练的皮肤属性评估模型得到所述目标用户皮肤所对应的至少一个第一维度属性值，所述皮肤属性评估模型是根据包含有第一维度属性值的图像训练得到的。
  12. 根据权利要求11所述的装置，所述装置还包括训练模块；
    所述训练模块，设置为将包含有第一维度属性的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性，并计算得到所述皮肤指标属性的实际值；根据所述实际值与参考值确定损失函数值；以及根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值；
    或，所述训练模块，设置为将包含有皮肤属性指标的样本图像输入到包含有权重参数初始值的皮肤属性评估模型中；通过所述皮肤属性评估模型提取所述样本图像中的皮肤指标属性；识别所述样本图像中皮肤指标属性所对应的样本区域；根据所述皮肤指标属性及对应的样本区域计算得到所述皮肤指标属性的实际值；根据所述实际值与参考值确定损失函数值；以及根据所述损失函数值调整所述皮肤属性评估模型的权重参数初始值，直到所述损失函数值满足预设条件时，得到所述皮肤属性评估模型的权重参数目标值。
  13. 根据权利要求10所述的装置，所述装置还包括归一化模块；
    所述归一化模块设置为对所述产品信息集合中的每个护肤品的第二维度属性总值进行归一化处理；
    所述归一化模块还设置为从所述产品信息集合中的每个护肤品的第二维度属性总值中选取最高值或最低值作为基准值，根据所述基准值对每个护肤品的第二维度属性总值进行归一化处理；
    或者，所述归一化模块还设置为将所述产品信息集合中的每个护肤品的第二维度属性总值的最高值对应到预设范围的上限值，将所述产品信息集合中的每个护肤品的第二维度属性总值的最低值对应到预设范围的下限值，再将位于所述产品信息集合中的每个护肤品的第二维度属性总值的最高值和最低值之间的总值对应到所述预设范围内。
  14. 根据权利要求10所述的装置，所述第一处理模块包括：
    系数获取单元，设置为获取表征第一维度属性和第二维度属性之间相关性关系的相关性系数；
    第一计算单元，设置为根据所述目标用户对应的每个第一维度属性值及对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值；以及
    第二计算单元，设置为根据所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值得到所述目标用户的至少一个第二维度属性值。
  15. 根据权利要求14所述的装置，所述第一计算单元还设置为根据所述目标用户对应的每个第一维度属性值、每个第一维度属性的权重系数和对应相关性系数得到所述目标用户的每个第二维度属性所对应的每个第一维度属性的组合值；或所述第一计算单元还设置为将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；根据所述目标用户对应的第一维度属性的需求等级值和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值；
    所述第一计算单元还设置为将所述目标用户对应的每个第一维度属性值转换为所述目标用户对应的每个第一维度属性的需求等级值；根据所述目标用户对应的第一维度属性的需求等级值、第一维度属性的权重系数和对应相关性系数的乘积得到所述目标用户对应的每个第二维度属性所对应的第一维度属性的组合值。
  16. 根据权利要求10所述的装置，所述第二处理模块还设置为将所述目标用户对应的第二维度属性值作为每个护肤品中对应的第二维度属性值，求取每个护肤品中全部的第二维度属性值之和，得到每个护肤品的第二维度属性总值。
  17. 根据权利要求10所述的装置，所述装置还包括：
    筛选信息获取模块，设置为获取产品筛选信息；
    筛选模块，设置为根据所述产品筛选信息从产品信息集合中筛选得到候选护肤品；
    所述第二处理模块还设置为根据所述目标用户对应的至少一个第二维度属性值和所述候选护肤品 中的每个护肤品的第二维度属性确定所述候选护肤品中的每个护肤品的第二维度属性总值。
  18. 根据权利要求10所述的装置，所述装置还包括显示模块和推送模块中至少一个；
    所述显示模块，设置为显示所述目标用户对应的待推送护肤品的全部或部分属性信息；
    所述推送模块，设置为推送所述目标用户对应的待推送护肤品的全部或部分属性信息。
  19. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现权利要求1至9中任一项所述方法的步骤。
  20. 一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现权利要求1至9中任一项所述的方法的步骤。
PCT/CN2020/071631 2019-01-31 2020-01-13 产品信息处理方法、装置、计算机设备和存储介质 WO2020156109A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910099040.9 2019-01-31
CN201910099040.9A CN111161007A (zh) 2019-01-31 2019-01-31 产品信息处理方法、装置、计算机设备和存储介质

Publications (1)

Publication Number Publication Date
WO2020156109A1 true WO2020156109A1 (zh) 2020-08-06

Family

ID=70555705

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/071631 WO2020156109A1 (zh) 2019-01-31 2020-01-13 产品信息处理方法、装置、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN111161007A (zh)
WO (1) WO2020156109A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113723984A (zh) * 2021-03-03 2021-11-30 京东城市(北京)数字科技有限公司 人群消费画像信息的获取方法、装置和存储介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112464706A (zh) * 2020-10-14 2021-03-09 鲁班嫡系机器人（深圳）有限公司 果实筛选、分拣方法、装置、系统、存储介质及设备
CN113742573B (zh) * 2021-08-03 2023-11-14 深圳Tcl新技术有限公司 一种多媒体资源搜索方法、装置、计算机设备和存储介质
CN115983928A (zh) * 2021-10-14 2023-04-18 北京字节跳动网络技术有限公司 作用于脸部的物品的确定方法、装置、设备和介质
CN117351545B (zh) * 2023-10-11 2024-03-15 广东圣千科技有限公司 一种基于大数据挖掘的护肤管理方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070255589A1 (en) * 2006-04-27 2007-11-01 Klinger Advanced Aesthetics, Inc. Systems and methods using a dynamic database to provide aesthetic improvement procedures
CN105662348A (zh) * 2016-01-11 2016-06-15 中山德尚伟业生物科技有限公司 基于智能手机的新型皮肤检测系统及产品辅助选择系统
CN108335727A (zh) * 2018-01-29 2018-07-27 杭州美界科技有限公司 一种基于历史记录的脸部护肤品推荐方法
CN108334589A (zh) * 2018-01-29 2018-07-27 杭州美界科技有限公司 一种脸部护肤品推荐方法
CN108364207A (zh) * 2018-01-29 2018-08-03 杭州美界科技有限公司 一种脸部护肤品及护肤方案推荐方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108198043A (zh) * 2018-01-29 2018-06-22 杭州美界科技有限公司 一种基于用户推荐的脸部护肤品推荐方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070255589A1 (en) * 2006-04-27 2007-11-01 Klinger Advanced Aesthetics, Inc. Systems and methods using a dynamic database to provide aesthetic improvement procedures
CN105662348A (zh) * 2016-01-11 2016-06-15 中山德尚伟业生物科技有限公司 基于智能手机的新型皮肤检测系统及产品辅助选择系统
CN108335727A (zh) * 2018-01-29 2018-07-27 杭州美界科技有限公司 一种基于历史记录的脸部护肤品推荐方法
CN108334589A (zh) * 2018-01-29 2018-07-27 杭州美界科技有限公司 一种脸部护肤品推荐方法
CN108364207A (zh) * 2018-01-29 2018-08-03 杭州美界科技有限公司 一种脸部护肤品及护肤方案推荐方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113723984A (zh) * 2021-03-03 2021-11-30 京东城市(北京)数字科技有限公司 人群消费画像信息的获取方法、装置和存储介质

Also Published As

Publication number Publication date
CN111161007A (zh) 2020-05-15

Similar Documents

Publication Publication Date Title
WO2020156109A1 (zh) 产品信息处理方法、装置、计算机设备和存储介质
CN110363387B (zh) 基于大数据的画像分析方法、装置、计算机设备及存储介质
Hunt et al. A general procedure for constructing mortality models
US9177230B2 (en) Demographic analysis of facial landmarks
JP7248820B2 (ja) 美容的皮膚属性を決定するための機器及び方法
Urbanová et al. Testing the reliability of software tools in sex and ancestry estimation in a multi-ancestral Brazilian sample
CN109801320B (zh) 一种基于面部分区的皮肤干燥状态智能识别方法及系统
KR102360993B1 (ko) 사용자에게 화장품을 추천하는 서비스를 제공하는 방법 및 장치
Jiang et al. On visual BMI analysis from facial images
Wadhwa et al. Age and Gender Determination from Finger Prints using RVA and dct coefficients
CN110751171A (zh) 图像数据分类方法、装置、计算机设备和存储介质
CN111954250A (zh) 一种轻量级Wi-Fi行为感知方法和系统
KR102450424B1 (ko) 사용자의 피부 타입과 화장품 성분에 따른 효능에 기반하여 사용자에게 적합한 화장품을 추천하는 방법 및 장치
Safira et al. K-Nearest Neighbour Classification and Feature Extraction GLCM for Identification of Terry's Nail
WO2021139112A1 (zh) 数据降维处理方法、装置、计算机设备和存储介质
Pérez-Bueno et al. A TV-based image processing framework for blind color deconvolution and classification of histological images
Messadi et al. Extraction of specific parameters for skin tumour classification
Thamizhvani et al. Mobile application-based computer-aided diagnosis of skin tumours from dermal images
KR20220005413A (ko) 멀티태스크 딥러닝을 이용한 지진 이벤트 분류 방법 및 장치
Nair et al. Non verbal behaviour analysis for distress detection using texture analysis
Rew et al. Monitoring skin condition using life activities on the SNS user documents
WO2023054008A1 (ja) 毛による肌と見た目の老化評価法
CN110840459A (zh) 人体平衡能力获取方法及系统、计算机设备及介质
Deshmukh et al. Human face aging based on deep learning: a survey
Bizel et al. Machine Learning Study: Identification of Skin Diseases for Various Skin Types Using Image Classification.

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20749486

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20749486

Country of ref document: EP

Kind code of ref document: A1