Cosmetic friction resistance evaluation method

Abstract

The invention relates to the technical field of cosmetics, and discloses a cosmetic friction resistance evaluation method which comprises the following steps: step 101, collecting cosmetic information; step 102, establishing an effect image generation model, and generating an effect image that the lipstick is applied to the lip of the user after six hours of the lipstick applying time selected by the user; step 103, based on the lipstick effect image generated by the effect image generation model, calculating a first coefficient of the percentage of the residual lipstick area of the lipstick corresponding to the lipstick effect image and a second coefficient of the friction resistance degree score; and step 104, evaluating the friction resistance of the lipstick according to the first coefficient and the second coefficient calculated in the step 103. According to the method, the lipstick formula information, the user lip information and the user behavior information are collected, the effect image generation model and the deep learning technology are combined, and the residual lipstick area proportion and the friction resistance degree score of the lipstick effect image are calculated, so that the friction resistance of the lipstick is evaluated.

Description

Technical Field

[0001] This invention relates to the field of cosmetics, and more specifically, to a method for evaluating the abrasion resistance of cosmetics. Background Technology

[0002] Cosmetics are any substance or mixture used by applying, spraying, or other similar methods to alter appearance or enhance attractiveness. Cosmetics mainly include lipstick, foundation, eyeshadow, mascara, and nail polish, used for beautifying, modifying, and caring for the skin. Lipstick is the most common type of cosmetic product.

[0003] Current lipstick product evaluation methods have several problems: traditional lipstick evaluation methods usually rely on manual observation and subjective judgment, and the evaluation results are easily affected by individual subjective factors, lacking objectivity and consistency. In addition, they require a large investment of manpower and time, which cannot meet the needs of rapid evaluation. This restricts the research and development and quality control of lipstick products. Existing evaluation methods mainly evaluate the appearance of lipstick and its appearance after application, without taking into account factors such as different lipstick formulas, different user lips, and changes in user behavior after application. This limits the accuracy and comprehensiveness of the evaluation results, and there is a lack of scientific and effective evaluation indicators and standards, making it difficult to achieve accurate comparison and evaluation of lipstick product performance. Summary of the Invention

[0004] This invention provides a method and system for evaluating the abrasion resistance of cosmetics to solve the aforementioned technical problems.

[0005] This invention provides a method for evaluating the abrasion resistance of cosmetics, comprising the following steps:

[0006] Step 101: Collect cosmetic information, including lipstick formula information, user lip information, and user behavior information;

[0007] Step 102: Establish an effect image generation model. Based on the effect image generation model, generate an effect image of the user's lips applying the lipstick six hours after the user selects the time to apply the lipstick. Input the lipstick formula information, user lip information, user behavior information, and the image of the user's lips into the effect image generation model, and output the effect image of the user's lips applying the lipstick six hours after the user selects the time to apply the lipstick.

[0008] Step 103: Based on the lipstick effect image generated by the effect image generation model, calculate the first coefficient of the percentage of residual lipstick area and the second coefficient of the friction resistance score of the lipstick effect image.

[0009] Step 104: Evaluate the lipstick's abrasion resistance based on the first and second coefficients calculated in Step 103.

[0010] In a preferred embodiment, in step 101, the lipstick formula information includes the types of pigments, oils, and moisturizing ingredients, and the content of pigments, oils, and moisturizing ingredients; the user lip information includes lip contour information, lip thickness information, lip moisture level, and lip texture; and the user behavior information indicates the time the user chooses to apply the lipstick.

[0011] In a preferred embodiment, in step 102, the lipstick formula information, user lip information, and user behavior information are encoded using one-hot encoding, bag-of-words model, or TF-IDF to obtain lipstick formula information feature vectors, user lip information feature vectors, and user behavior information feature vectors.

[0012] In a preferred embodiment, in step 102, the effect image generation model includes: a combination layer, a hidden layer, and a decoding layer. The combination layer is used to combine the lipstick formula information feature vector, the user lip information feature vector, and the user behavior information feature vector to obtain a combination feature. The hidden layer takes the combination feature as input and outputs a color change matrix to the decoding layer. The decoding layer takes the color change matrix and the user lip image as input and outputs an effect image of the user's lips six hours after the user selects the time to apply the lipstick.

[0013] The color variation matrix is ​​the same size as the user's lip image;

[0014] The formula for calculating the composite layer is:

[0015] Z = σ(Z1||Z2||Z3)

[0016] Where Z represents the combined feature, Z1, Z2 and Z3 represent the lipstick formula information feature vector, the user lip information feature vector and the user behavior information feature vector, respectively, and σ represents the sigmoid function;

[0017] The formula for calculating the hidden layer is:

[0018] S r =MLP1(Z)

[0019] S g =MLP2(Z)

[0020] S b =MLP3(Z)

[0021] S r S g S bThe first, second, and third color change vectors are represented by M dividing vectors into M partition vectors, each partition vector having M components. The M partition vectors are then concatenated to form a color change matrix. The first, second, and third color change vectors are concatenated to form the first, second, and third color change matrices, respectively. MLP1, MLP2, and MLP3 represent inputting Z into the first, second, and third multilayer perceptrons for processing, respectively.

[0022] The calculation formula for the decoding layer is:

[0023] P r =J r ◎X r

[0024] P g =J g ◎X g

[0025] P b =J b ◎X b

[0026] P r P g P b J represents the r, g, and b color channels of the output image, respectively. r J g J b These represent the first, second, and third color change matrices, respectively. ◎ indicates element-wise multiplication at corresponding positions in the matrix. Xr, X g X b These represent the r, g, and b color channels of the input user's lip image, respectively.

[0027] In a preferred embodiment, the hidden layer includes multiple time steps. After the color change matrix output at the t-th time step is input into the decoding layer, the result is an image of the user's lips after the t-th hour since the user selected the time to apply the lipstick.

[0028] The formula for calculating the hidden layer is as follows:

[0029] H (0) =MLP(Z)

[0030] u (t) =σ(W u MLP(Z)+W u H (t-1) +b u )

[0031] r (t) =σ(W r MLP(Z)+W r H (t-1)+b r )

[0032] C (t) =tanh(W c MLP(Z)+W c r (t) ⊙H (t-1) +b c )

[0033] H (t) =(1-u (t) )⊙C (t) +u (t) ⊙H (t-1)

[0034] W u W r W c W j b represents the weight parameter. u b r b c Let Z represent the bias parameter, ⊙ represent the dot product, and Z represent the combined features. (t) and H (t-1) These represent the outputs at the t-th and (t-1)-th time steps, respectively.

[0035] MLP(Z) involves inputting Z into a first, second, and third multilayer perceptron for processing. Each time step contains the three branches of the aforementioned hidden layer, and the output of each time step contains the first, second, and third color change vectors, respectively.

[0036] In a preferred embodiment, the method for calculating the percentage of the residual lipstick area corresponding to the lipstick effect image as a first coefficient in step 103 includes the following steps:

[0037] Step 1: Based on the image state at the time when the user selects the lipstick to be applied to the user's lips, the upper limit of the lip color value in the RGB color space is set, and the lower limit of the lip color value in the RGB color space is set in the image state when the user does not apply lipstick. Set the color threshold range.

[0038] Step 2: Based on the color threshold range, mark the pixels in the lip area of ​​the lip effect image that match the color range as pixels in the lip area, and mark the area outside the lip area in the lip area as non-lip area.

[0039] Step 3: Count the number of pixels N in all lipstick areas and the number of pixels M in the lip area;

[0040] Step 4: Calculate the percentage of residual lipstick area in the final image, using the first parameter.

[0041] The formula for calculating the percentage of residual lipstick area is as follows:

[0042] K b = (N / M) * 100%

[0043] Among them, K b This represents the percentage of residual lipstick area in the effect image. N represents the number of pixels in the lipstick area of ​​the effect image, and M represents the number of pixels in the lip area of ​​the effect image. Both N and M are integers.

[0044] In a preferred embodiment, the method for calculating the second coefficient of the friction resistance score for all corresponding lipstick areas in the lipstick effect image in step 103 includes the following steps:

[0045] Step 1: Use an image segmentation algorithm to segment the lipstick-preserving area within the upper and lower lip regions of the lipstick effect image from the other parts of the upper and lower lip regions;

[0046] Step 2: Count the number of pixels in the lipstick retention area and the number of pixels in the upper and lower lip areas of the lipstick effect image. Use nested loops to traverse all pixels in the lipstick retention area and obtain the RGB values ​​of all pixels. Accumulate the values ​​of the three color channels of all pixels and calculate the first average value of the three color channels of the pixels in the lipstick retention area. The first average value represents the average color of the lipstick retention area.

[0047] Step 3: Use nested loops to iterate through the RGB values ​​of all pixels in the lip area at the time the user selects lipstick to apply it to the user's lips. Calculate the second average of the three color channels of the pixels in the lip area at the time the user selects lipstick to apply it to the user's lips. The second average represents the color of the lip area at the time the user selects lipstick to apply it to the user's lips.

[0048] Step 4: Calculate the difference between the R color channel values ​​of the first average value and the second average value. The smaller the difference, the better the friction resistance; the larger the difference, the worse the friction resistance.

[0049] In a preferred embodiment, in step 104, the formula for calculating the lipstick's abrasion resistance is:

[0050]

[0051] Among them, C d K indicates the lipstick's abrasion resistance. b The first coefficient represents the percentage of residual lipstick area in the final image. ω1 and ω2 represent the difference between the R color channel values ​​of the first average and the second average, respectively, and are random numbers in [0, 1]. e represents the natural exponent.

[0052] A cosmetic abrasion resistance evaluation system, characterized in that it comprises:

[0053] The information collection module is used to collect lipstick formula information, user lip information, and user behavior information;

[0054] The effect image generation module is used to generate an effect image of the user's lips six hours after the user selects when to apply the lipstick. The input of lipstick formula information, user lip information, user behavior information, and an image of the user's lips is fed into the effect image generation model, and the output is an effect image of the user's lips six hours after the user selects when to apply the lipstick.

[0055] The first coefficient calculation module is used to calculate the percentage of residual lipstick area corresponding to the lipstick effect image.

[0056] The second coefficient calculation module is used to calculate the second coefficient of the friction resistance score corresponding to the lipstick effect image;

[0057] The friction resistance evaluation module evaluates the friction resistance of lipsticks based on the first and second coefficients.

[0058] A storage medium storing non-transitory computer-readable instructions, which, when executed by a computer, can perform the steps in a cosmetic abrasion resistance evaluation method as described above.

[0059] The beneficial effects of this invention are as follows: By collecting lipstick formula information, user lip information, and user behavior information, and combining effect image generation models and deep learning technology, this invention calculates the residual lipstick area ratio and friction resistance score of the lipstick effect image, thereby evaluating the friction resistance performance of the lipstick. This provides a scientific and effective evaluation method for the research and development and quality control of lipstick products, and helps to improve the quality and durability of lipstick products. Attached Figure Description

[0060] Figure 1 This is a flowchart of the cosmetic friction resistance evaluation method of the present invention.

[0061] Figure 2 This is a block diagram of the cosmetic friction resistance evaluation system of the present invention.

[0062] In the diagram: 100, Information Acquisition Module; 200, Effect Image Generation Module; 300, First Coefficient Calculation Module; 400, Second Coefficient Calculation Module; 500, Friction Resistance Evaluation Module. Detailed Implementation

[0063] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0064] like Figure 1 As shown, a method for evaluating the abrasion resistance of cosmetics includes the following steps:

[0065] Step 101: Collect cosmetic information, including lipstick formula information, user lip information, and user behavior information;

[0066] Lipstick formula information includes the types of pigments, oils, and moisturizing ingredients, as well as their concentrations.

[0067] Pigments include synthetic and natural pigments; oils include plant and animal oils; and moisturizing ingredients include skin oils, plant oils, natural beeswax, seaweed extracts, and vitamin E. The content of pigments, oils, and moisturizing ingredients in the lipstick formula information is expressed as a weight ratio. For example, a lipstick formula includes 10-15 parts pigment, 65-75 parts oil, 18-24 parts moisturizing ingredients, and 26-38 parts other components.

[0068] User lip information includes lip contour information, lip thickness information, lip moisture level, and lip texture;

[0069] The method for obtaining user lip information is as follows:

[0070] The system captures images of a user's lips with lipstick applied, and uses face detection and lip detection algorithms from image processing technology to locate the upper and lower lip regions and segment them from the image.

[0071] The thickness information of the lips is obtained by establishing a two-dimensional coordinate system with the lower left corner of the user's lip image as the origin, obtaining the outer contour line of the upper and lower lip regions in the user's lip image, finding the maximum and minimum values ​​of the ordinate on the outer contour line, and calculating the difference between the maximum and minimum values ​​of the ordinate. This difference represents the thickness of the lips.

[0072] The formula for calculating lip thickness is:

[0073] T lip =|y max -y min |

[0074] Among them, Tlip Indicates the thickness of the lips, y max The y-coordinate represents the maximum value of the vertical axis along the outer contour line of the upper and lower lip regions. min This represents the minimum vertical coordinate value on the outer contour line of the upper and lower lip regions;

[0075] The contour information of the lips is represented by the area of ​​the upper and lower lip regions in the user's lip image. A two-dimensional coordinate system is established with the lower left corner of the user's lip image as the origin of the coordinate system. The outer contour lines that make up the upper and lower lip regions in the user's lip image are obtained, and the maximum and minimum values ​​of the ordinates and the maximum and minimum values ​​of the abscissas on the outer contour lines are found.

[0076] The formula for calculating the area of ​​the upper and lower lip regions is:

[0077]

[0078] Where ΔA represents the area of ​​the upper and lower lip regions, y max The y-coordinate represents the maximum value of the vertical axis along the outer contour line of the upper and lower lip regions. min The x-coordinate represents the minimum value of the vertical axis along the outer contour line of the upper and lower lip regions. max The x-coordinate represents the maximum value of the x-coordinate on the outer contour line of the upper and lower lip regions. min This represents the minimum x-coordinate value on the outer contour line of the upper and lower lip regions;

[0079] The moisture level of the lips is represented by the average gray value of the pixels in the upper and lower lip regions of the user's lip image. The gray value is in the range of [0, 255]. A smaller gray value indicates poor moisture level, while a larger gray value indicates high moisture level. The average gray value of all pixels in the upper and lower lip regions of the user's lip image is calculated by obtaining all pixels in the region.

[0080] The formula for calculating the moisture level of the lips is:

[0081]

[0082] Among them, M os Indicates the degree of moisture of the lips, with a value range of [0, 1]. os A higher value indicates a higher level of lip moisture, and vice versa. c Let p represent the set of all pixels within the upper and lower lip regions of the user's lip image, and let p represent the number of pixels within those regions. i σ represents the grayscale value of the i-th pixel within the upper and lower lip regions of the user's lip image, and σ is a sigmoid function;

[0083] The texture of the lips is represented by the total length of the lip lines in the image;

[0084] The total length of lip lines can be calculated using the following steps:

[0085] Step 1: Within the lip area, use edge detection algorithms or texture feature extraction algorithms to identify the lip lines, thereby extracting the contour information of all lines within the lip area.

[0086] Step 2: Calculate the length of all lip lines by extracting them, and add them together to get the total length;

[0087] The formula for calculating the total length of lip lines is as follows:

[0088]

[0089]

[0090] Where L represents the total length of the lip lines, m represents the number of coordinate points on the line outline, and P j and P j+1 Let d(P) represent the j-th and (j+1)-th coordinates on the texture outline, respectively. j P j+1 () represents the Euclidean distance between the j-th and (j+1)-th coordinate points;

[0091] For example, in step two, curve length calculation functions such as OpenCV and Matlab can be used to calculate the length of each texture segment; the texture outline is converted into a sequence of coordinate points, and then the length of the texture segment is estimated by calculating the distance between adjacent points, and they are summed up to obtain the total length;

[0092] Step 3: Smooth the calculation results from Step 2 or remove interfering terms to obtain a more accurate total length of lip lines;

[0093] Step 4: Convert the total length of the lip lines from Step 3 into units such as pixels or millimeters to represent the total length of the lip lines.

[0094] User behavior information indicates the time when the user chooses to apply lipstick, which is a time point randomly selected by the user within the range of 6:00 AM to 10:00 PM.

[0095] Step 102: Establish an effect image generation model, and generate an effect image of the user's lips 6 hours after the user selects the time to apply the lipstick.

[0096] In one embodiment of the present invention, lipstick formula information, user lip information, user behavior information, and an image of the user's lips are input into the effect image generation model, and the effect image of the user's lips after 6 hours of the time when the user selected to apply the lipstick is output.

[0097] Encode lipstick formula information, user lip information, and user behavior information to obtain lipstick formula information feature vector, user lip information feature vector, and user behavior information feature vector;

[0098] In one embodiment of the present invention, the encoding method is one-hot encoding, bag of words (BoW), or TF-IDF (Term Frequency-Inverse Document Frequency).

[0099] The effect image generation model includes: a combination layer, a hidden layer, and a decoding layer. The combination layer is used to combine the lipstick formula information feature vector, the user's lip information feature vector, and the user's behavior information feature vector to obtain the combined features. The hidden layer takes the combined features as input and outputs the color change matrix to the decoding layer. The decoding layer takes the color change matrix and the user's lip image as input and outputs the effect image of the user's lips 6 hours after the user selects the time to apply the lipstick.

[0100] The color variation matrix is ​​the same size as the user's lip image;

[0101] The formula for calculating the composite layer is:

[0102] Z = σ(Z1||Z2||Z3)

[0103] Where Z represents the combined feature, Z1, Z2 and Z3 represent the lipstick formula information feature vector, the user lip information feature vector and the user behavior information feature vector, respectively, and σ represents the sigmoid function.

[0104] The formula for calculating the hidden layer is:

[0105] S r =MLP1(Z)

[0106] S g =MLP2(Z)

[0107] S b =MLP3(Z)

[0108] S r S g S bThe first, second, and third color change vectors are represented. The color change vectors are divided into M equal partition vectors, with M components in each partition vector. The M partition vectors are then concatenated to form a color change matrix. The first, second, and third color change vectors are concatenated to form the first, second, and third color change matrices, respectively. MLP1, MLP2, and MLP3 represent inputting Z into the first, second, and third multilayer perceptrons for processing, respectively.

[0109] The calculation formula for the decoding layer is:

[0110] P r =J r ◎X r

[0111] P g =J g ◎X g

[0112] P b =J b ◎X b

[0113] P r P g P b J represents the r, g, and b color channels of the output image, respectively. r J g J b These represent the first, second, and third color change matrices, respectively. ◎ indicates element-wise multiplication at corresponding positions in the matrix. Xr, X g X b These represent the r, g, and b color channels of the input user's lip image, respectively.

[0114] In one embodiment of the present invention, the hidden layer includes multiple time steps. After the color change matrix output at the t-th time step is input into the decoding layer, the result is an image of the user's lips after the t-th hour since the user selected the time to apply the lipstick.

[0115] The formula for calculating the hidden layer is as follows:

[0116] H (0) =MLP(Z)

[0117] u (t) =σ(W u MLP(Z)+W u H (t-1) +b u )

[0118] r (t) =σ(W r MLP(Z)+W r H (t-1)+b r )

[0119] C (t) =tanh(W c MLP(Z)+W c r (t) ⊙H (t-1) +b c )

[0120] H (t) =(1-u (t) )⊙C (t) +u (t) ⊙H (t-1)

[0121] W u W r W c W j b represents the weight parameter. u b r b c Let Z represent the bias parameter, ⊙ represent the dot product, and Z represent the combined features. (t) and H (t-1) These represent the outputs at the t-th and (t-1)-th time steps, respectively.

[0122] MLP(Z) involves inputting Z into a first, second, and third multilayer perceptron for processing. Each time step contains the three branches of the aforementioned hidden layer, and the output of each time step contains the first, second, and third color change vectors, respectively.

[0123] Step 103: Based on the lipstick effect image generated by the effect image generation model, calculate the first coefficient of the percentage of residual lipstick area and the second coefficient of the friction resistance score of the lipstick effect image.

[0124] It should be noted that the abrasion resistance score is represented by the average RGB value of the pixels in the lipstick-preserved area of ​​the lipstick effect image.

[0125] In one embodiment of the present invention, a method for calculating a first coefficient representing the percentage of residual lipstick area corresponding to a lipstick effect image is provided, comprising the following steps:

[0126] Step 1: Based on the image state at the time when the user selects the lipstick to be applied to the user's lips, the upper limit of the lip color value in the RGB color space is set, and the lower limit of the lip color value in the RGB color space is set in the image state when the user does not apply lipstick. Set the color threshold range.

[0127] For example, the color range of the lip area in images of a user with and without lipstick is (150, 0, 0) - (255, 100, 100);

[0128] Step 2: Preprocess the lipstick effect image generated by the effect image generation model, convert it to RGB color space, and then mark the pixels in the lip area of ​​the lip effect image that meet the color range as lipstick area pixels according to the color threshold range, and mark the other pixels in the lip area of ​​the lip effect image as non-lipstick area pixels.

[0129] Step 3: Count the number of pixels N in all lipstick areas and the number of other pixels M in the lip area of ​​the lipstick effect image;

[0130] Step 4: Calculate the percentage of residual lipstick area in the final image;

[0131] The formula for calculating the percentage of residual lipstick area is as follows:

[0132] K b = (N / M) * 100%

[0133] Among them, K b This represents the percentage of residual lipstick area in the effect image. N represents the number of pixels in the lipstick area of ​​the effect image, and M represents the number of pixels in the lip area of ​​the effect image. Both N and M are integers.

[0134] In one embodiment of the present invention, a method for calculating the second coefficient of the friction resistance score for all corresponding lipstick areas in a lipstick effect image is provided, comprising the following steps:

[0135] Step 1: Use an image segmentation algorithm to separate the lipstick-preserving area from the other parts of the upper and lower lip regions in the lipstick effect image; specifically, threshold segmentation and edge detection algorithms can be used.

[0136] Step 2: Count the number of pixels in the lipstick retention area and the number of pixels in the upper and lower lip areas of the lipstick effect image. Use nested loops to traverse all pixels in the lipstick retention area and obtain the RGB values ​​of all pixels. Accumulate the values ​​of the three color channels (R, G, B) of all pixels and calculate the first average value of the three color channels (R, G, B) of the pixels in the lipstick retention area. The first average value represents the average color of the lipstick retention area.

[0137] Step 3: Repeat Step 1 and Step 2, using nested loops to iterate through the RGB values ​​of all pixels in the lip area at the time the user's selected lipstick is applied to the user's lips. Calculate the second average of the three color channels (R, G, B) of the pixels in the lip area at the time the image is applied. The second average represents the color of the lip area at the time the image is applied to the user's lips.

[0138] Step 4: Calculate the difference between the R color channel values ​​of the first average value and the second average value. The smaller the difference, the better the friction resistance; the larger the difference, the worse the friction resistance.

[0139] Step 104: Evaluate the lipstick's abrasion resistance based on the first and second coefficients calculated in Step 103;

[0140] In one embodiment of the present invention, the formula for evaluating the abrasion resistance of lipstick is as follows:

[0141]

[0142] Among them, C d K indicates the lipstick's abrasion resistance. b The first coefficient represents the percentage of residual lipstick area in the final image. ω1 and ω2 represent the difference between the R color channel values ​​of the first average and the second average, respectively, and are random numbers in [0,1]. ω1 and ω2 are determined by cosmetic engineers based on design standards or design requirements, and e represents the natural index.

[0143] Design a friction resistance evaluation range of [0,1]. This range can be designed as four levels: Level 3: [0,0.25], Level 2: [0.25,0.5], Level 1: [0.5,0.75], and Special Level: [0.75,1]. If the calculated friction resistance value is 0.86, it means that the corresponding lipstick belongs to the range of [0.75,1], and the friction resistance of this lipstick is Special Level.

[0144] like Figure 2 As shown, a cosmetic friction resistance evaluation system includes the following modules:

[0145] The information collection module 100 is used to collect lipstick formula information, user lip information, and user behavior information.

[0146] The effect image generation module 200 is used to generate an effect image of the user's lips applying the lipstick 6 hours after the user selects the time to apply the lipstick. It inputs lipstick formula information, user lip information, user behavior information, and an image of the user's lips into the effect image generation model and outputs an effect image of the user's lips applying the lipstick 6 hours after the user selects the time to apply the lipstick.

[0147] The first coefficient calculation module 300 is used to calculate the percentage of residual lipstick area corresponding to the lipstick effect image as the first coefficient.

[0148] The second coefficient calculation module 400 is used to calculate the second coefficient of the friction resistance score corresponding to the lipstick effect image;

[0149] The friction resistance evaluation module 500 evaluates the friction resistance of lipsticks based on the first and second coefficients.

[0150] A storage medium storing non-transitory computer-readable instructions, which, when executed by a computer, can perform the steps in a cosmetic abrasion resistance evaluation method as described above.

[0151] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A method for evaluating the abrasion resistance of cosmetics, characterized in that, Includes the following steps: Step 101: Collect cosmetic information, including lipstick formula information, user lip information, and user behavior information; Step 102: Establish an effect image generation model. Based on the effect image generation model, generate an effect image of the user's lips applying the lipstick six hours after the user selects the time to apply the lipstick. Input the lipstick formula information, user lip information, user behavior information, and the image of the user's lips into the effect image generation model, and output the effect image of the user's lips applying the lipstick six hours after the user selects the time to apply the lipstick. Step 103: Based on the lipstick effect image generated by the effect image generation model, calculate the first coefficient of the percentage of residual lipstick area and the second coefficient of the friction resistance score of the lipstick effect image. Step 104: Evaluate the lipstick's abrasion resistance based on the first and second coefficients calculated in Step 103.

2. The method for evaluating the abrasion resistance of cosmetics according to claim 1, characterized in that, In step 101, the lipstick formula information includes the types of pigments, oils, and moisturizing ingredients, and the content of pigments, oils, and moisturizing ingredients; the user lip information includes lip contour information, lip thickness information, lip moisture level, and lip texture; and the user behavior information indicates the time the user chooses to apply the lipstick.

3. The method for evaluating the abrasion resistance of cosmetics according to claim 2, characterized in that, In step 102, the lipstick formula information, user lip information, and user behavior information are encoded using one-hot encoding, bag-of-words model, or TF-IDF to obtain lipstick formula information feature vector, user lip information feature vector, and user behavior information feature vector.

4. The method for evaluating the abrasion resistance of cosmetics according to claim 3, characterized in that, In step 102, the effect image generation model includes: a combination layer, a hidden layer, and a decoding layer. The combination layer is used to combine the lipstick formula information feature vector, the user's lip information feature vector, and the user's behavior information feature vector to obtain the combined features. The hidden layer takes the combined features as input and outputs a color change matrix to the decoding layer. The decoding layer takes the color change matrix and the user's lip image as input and outputs an effect image of the user's lips six hours after the user selects the time to apply the lipstick. The color variation matrix is ​​the same size as the user's lip image; The formula for calculating the composite layer is: Z = σ(Z1||Z2||Z3) Where Z represents the combined feature, Z1, Z2 and Z3 represent the lipstick formula information feature vector, the user lip information feature vector and the user behavior information feature vector, respectively, and σ represents the sigmoid function; The formula for calculating the hidden layer is: S r =MLP1(Z) S g =MLP2(Z) S b =MLP3(Z) S r S g S b The first, second, and third color change vectors are represented by M dividing vectors into M partition vectors, each partition vector having M components. The M partition vectors are then concatenated to form a color change matrix. The first, second, and third color change vectors are concatenated to form the first, second, and third color change matrices, respectively. MLP1, MLP2, and MLP3 represent inputting Z into the first, second, and third multilayer perceptrons for processing, respectively. The calculation formula for the decoding layer is: P r =J r ◎X r P g =J g ◎X g P b =J b ◎X b P r P g P b J represents the r, g, and b color channels of the output image, respectively. r J g J b These represent the first, second, and third color change matrices, respectively. ◎ indicates element-wise multiplication at corresponding positions in the matrix. X r X g X b These represent the r, g, and b color channels of the input user's lip image, respectively.

5. The method for evaluating the abrasion resistance of cosmetics according to claim 4, characterized in that, The hidden layer consists of multiple time steps. The color change matrix output at the t-th time step is input into the decoding layer to obtain the effect image of the user's lips after the t-th hour since the user selected the time to apply the lipstick. The formula for calculating the hidden layer is as follows: H (0) =MLP(Z) at (t) =σ(W u MLP(Z)+W u H (t-1) +b u ) r (t) =σ(W r MLP(Z)+W r H (t-1) +b r ) C (t) =tanh(W c MLP(Z)+W c r (t) ⊙H (t-1) +b c ) H (t) =(1-u (t) )⊙C (t) +u (t) ⊙H (t-1) W u W r W c W j b represents the weight parameter. u b r b c Let Z represent the bias parameter, ⊙ represent the dot product, and Z represent the combined features. (t) and H (t-1) These represent the outputs at the t-th and (t-1)-th time steps, respectively. MLP(Z) involves inputting Z into a first, second, and third multilayer perceptron for processing. Each time step contains the three branches of the aforementioned hidden layer, and the output of each time step contains the first, second, and third color change vectors, respectively.

6. The method for evaluating the abrasion resistance of cosmetics according to claim 5, characterized in that, In step 103, the calculation method for the first coefficient of the percentage of residual lipstick area corresponding to the lipstick effect image includes the following steps: Step 1: Based on the image state at the time when the user selects the lipstick to be applied to the user's lips, the upper limit of the lip color value in the RGB color space is set, and the lower limit of the lip color value in the RGB color space is set in the image state when the user does not apply lipstick. Set the color threshold range. Step 2: Based on the color threshold range, mark the pixels in the lip area of ​​the lip effect image that match the color range as pixels in the lip area, and mark the area outside the lip area in the lip area as non-lip area. Step 3: Count the number of pixels N in all lipstick areas and the number of pixels M in the lip area; Step 4: Calculate the percentage of residual lipstick area in the final image, using the first parameter. The formula for calculating the percentage of residual lipstick area is as follows: K b =(N / M)*100% Among them, K b This represents the percentage of residual lipstick area in the effect image. N represents the number of pixels in the lipstick area of ​​the effect image, and M represents the number of pixels in the lip area of ​​the effect image. Both N and M are integers.

7. The method for evaluating the abrasion resistance of cosmetics according to claim 6, characterized in that, In step 103, the method for calculating the second coefficient of the friction resistance score for all corresponding lipstick areas in the lipstick effect image includes the following steps: Step 1: Use an image segmentation algorithm to segment the lipstick-preserving area within the upper and lower lip regions of the lipstick effect image from the other parts of the upper and lower lip regions; Step 2: Count the number of pixels in the lipstick retention area and the number of pixels in the upper and lower lip areas of the lipstick effect image. Use nested loops to traverse all pixels in the lipstick retention area and obtain the RGB values ​​of all pixels. Accumulate the values ​​of the three color channels of all pixels and calculate the first average value of the three color channels of the pixels in the lipstick retention area. The first average value represents the average color of the lipstick retention area. Step 3: Use nested loops to iterate through the RGB values ​​of all pixels in the lip area at the time the user's selected lipstick is applied to the user's lips. Calculate the second average of the three color channels of the pixels in the lip area at the time the image is applied. The second average represents the color of the lip area at the time the image is applied to the user's lips. Step 4: Calculate the difference between the R color channel values ​​of the first average value and the second average value. The smaller the difference, the better the friction resistance; the larger the difference, the worse the friction resistance.

8. The method for evaluating the abrasion resistance of cosmetics according to claim 7, characterized in that, In step 104, the formula for evaluating the lipstick's abrasion resistance is as follows: Among them, C d K indicates the lipstick's abrasion resistance. b The first coefficient represents the percentage of residual lipstick area in the final image. ω1 and ω2 represent the difference between the R color channel values ​​of the first average and the second average, respectively, and are random numbers in [0,1]. e represents the natural exponent.

9. A cosmetic friction resistance evaluation system, characterized in that, include: The information collection module (100) is used to collect lipstick formula information, user lip information and user behavior information; The effect image generation module (200) is used to generate an effect image of the user's lips applying lipstick six hours after the user selects the time to apply lipstick. It inputs lipstick formula information, user lip information, user behavior information, and an image of the user's lips into the effect image generation model and outputs an effect image of the user's lips applying the lipstick six hours after the user selects the time to apply lipstick. The first coefficient calculation module (300) is used to calculate the percentage of the residual lipstick area corresponding to the lipstick effect image as the first coefficient. The second coefficient calculation module (400) is used to calculate the second coefficient of the friction resistance score corresponding to the lipstick effect image; The friction resistance evaluation module (500) evaluates the friction resistance of lipstick based on the first coefficient and the second coefficient.

10. A storage medium, characterized in that, It stores non-transitory computer-readable instructions, which, when executed by a computer, can perform the steps in the cosmetic abrasion resistance evaluation method as described in claim 9.

Images