Cosmetic soothing efficacy evaluation method and system based on deep learning model

Abstract

The invention discloses a cosmetic soothing efficacy evaluation method and system based on a deep learning model. The method comprises the following steps: setting a control group and a to-be-tested sample group; after the groups are cultured for the same time, fixing the zebrafish by using methyl cellulose, and shooting side views of the groups of zebrafish under a fluorescence microscope to serve as original zebrafish images; obtaining a binary mask according to the preprocessed original zebra fish image; taking the preprocessed zebra fish original image and the corresponding binary mask as samples in a training data set; training the semantic segmentation model by using the training data set; and preprocessing the zebra fish images to be detected in each group, inputting the zebra fish images to be detected in each group into the trained semantic segmentation model to obtain images of each group only containing the target area, and extracting the number of fluorescent particles in each group to evaluate the soothing effect of the cosmetics to be detected. According to the method, the fluorescence imaging technology and the deep learning model are combined, and the number of the fluorescent particles in the key area of the zebra fish image can be rapidly and accurately calculated, so that evaluation of the sustained-release effect of the cosmetics is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of cosmetic efficacy evaluation, and in particular, to a method, system, terminal device, computer-readable storage medium, and computer program product for evaluating the soothing efficacy of cosmetics based on a deep learning model. Background Art

[0002] The safety assessment of traditional cosmetics mainly relies on animal experiments, such as those using mice, rats, and rabbits. However, animal experiments not only have animal welfare and ethical issues but also have some limiting factors, such as the relatively long life cycle of animals, the difficulty of breeding offspring, and the relatively difficult experimental operations. In addition, animal experiments are easily affected by external factors such as constitution and environment, resulting in non-repeatability and untrustworthiness of the results.

[0003] As a biological model, zebrafish has characteristics such as transparency and a short breeding cycle, and is an ideal model for studying gene functions and drug screening. Early zebrafish imaging technology was mainly applied to research aspects such as embryonic development and kinematics. With the development of fluorescence imaging technology, zebrafish fluorescence imaging technology has gradually been applied to evaluate the effects of cosmetics, and its advantages are effectiveness, high throughput, and precision. The general process of using fluorescence imaging technology to detect the efficacy of cosmetics is as follows: Transfer normally developed 2dpf zebrafish into a six-well plate; incubate the zebrafish under different experimental conditions respectively; fix the zebrafish after incubation and take pictures under a fluorescence microscope; count the number of neutrophils according to the images. Although this method can evaluate the efficacy of cosmetics by counting the number of neutrophils, it mainly has the following deficiencies: 1) A large amount of image data is obtained in the experiment, and the manual counting task is heavy. Manual counting is extremely dependent on the energy and concentration of the operator. Prolonged screening on the screen is likely to cause visual fatigue, resulting in misjudgment and missed judgment. At the same time, due to the accumulation of fatigue, the counting accuracy will decline with the working hours, making it difficult to guarantee the data consistency between batches; 2) Manual counting has certain subjective limitations. It is difficult to unify the morphological judgment criteria for neutrophils by different operators, and this subjective difference will lead to large fluctuations in the counting results, resulting in a lack of credibility of the results; 3) Low efficiency. Manual counting can only identify each image and each cell one by one, with extremely low efficiency, resulting in a long cycle for evaluating the efficacy of cosmetics and delaying the product R & D progress. Summary of the Invention

[0004] In order to solve the above deficiencies of the prior art, the present invention provides a method, system, terminal device, computer-readable storage medium, and computer program product for evaluating the soothing efficacy of cosmetics based on a deep learning model. By combining fluorescence imaging technology and a deep learning model, the number of fluorescent particles in the key area of the zebrafish image can be quickly and accurately extracted, thereby realizing the evaluation of the soothing efficacy of cosmetics.

[0005] The first object of the present invention is to provide a method for evaluating the soothing effect of cosmetics based on a deep learning model.

[0006] The second object of the present invention is to provide a system for evaluating the soothing effect of cosmetics based on a deep learning model.

[0007] The third object of the present invention is to provide a terminal device.

[0008] The fourth object of the present invention is to provide a computer-readable storage medium.

[0009] The fifth object of the present invention is to provide a computer program product.

[0010] The first object of the present invention can be achieved by adopting the following technical solutions: A method for evaluating the soothing effect of cosmetics based on a deep learning model, the method comprising: Setting a normal control group: putting zebrafish into standard dilution water; Setting a model control group: putting zebrafish into standard dilution water containing a modeling agent; Setting a test sample group: putting zebrafish into standard dilution water containing a modeling agent and a test sample stock solution; the test sample stock solution is a solution prepared from a cosmetic to be tested; Setting a positive control group: putting zebrafish into standard dilution water containing a modeling agent and dipotassium glycyrrhizinate; The zebrafish in each group are 2dpf neutrophil gene mutant zebrafish with normal development; After culturing each group for the same time, fixing the zebrafish with methyl cellulose and taking a side view of the zebrafish in each group under a fluorescence microscope as the original zebrafish image; Preprocessing the original zebrafish image to improve the overall brightness of the image; According to the preprocessed original zebrafish image, obtaining a binary mask; using the preprocessed original zebrafish image and the corresponding binary mask as samples in the training dataset; Training a semantic segmentation model using the training dataset; Inputting the preprocessed zebrafish images to be detected in each group into the trained semantic segmentation model to obtain images in each group that only contain the target area; according to the images in each group that only contain the target area, extracting the number of fluorescent particles in each group; according to the number of fluorescent particles in each group, evaluating the soothing effect of the cosmetic to be tested.

[0011] Further, the step of inputting the preprocessed zebrafish images to be detected in each group into the trained semantic segmentation model to obtain images in each group that only contain the target area includes: After preprocessing the zebrafish images to be detected in each group, input them into the trained semantic segmentation model to output the corresponding binary mask; Perform pixel-level multiplication on the binary mask and the corresponding preprocessed zebrafish image to be detected to obtain a zebrafish image containing only the target area.

[0012] Further, the preprocessing includes: Convert the image from the RGB space to the HSV space; Increase the pixel values of the V channel and the S channel of the image to obtain an image with increased brightness; Convert the image with increased brightness from the HSV space to the RGB space; Wherein, the image is the original zebrafish image or the zebrafish images to be detected in each group.

[0013] Further, obtaining the binary mask according to the preprocessed original zebrafish image includes: Circle the target area in the preprocessed original zebrafish image with a thin line to obtain a preliminary line contour; the target area is two areas corresponding to the zebrafish yolk sac and the back above the lateral line; On the basis of the preliminary line contour, use morphological operations to refine the line contour to remove existing noise or artifacts; Finally, fill the line contour with color and binarize it to obtain the binary mask.

[0014] Further, extracting the number of fluorescent particles in each group according to the images containing only the target area in each group includes: Extract the green channel from the RGB channels of the image containing only the target area and convert it into a grayscale image; screen out the pixels in the grayscale image whose pixels are greater than the set threshold, and then use the SimpleITK toolkit to regard the connected pixels as a fluorescent particle, so as to obtain the number of fluorescent particles in the image.

[0015] Further, training the semantic segmentation model using the training dataset includes: Input the preprocessed original zebrafish image into the semantic segmentation model to output the binary mask; Compare the output binary mask with the binary mask corresponding to the preprocessed original zebrafish image, calculate the difference value between the two using the cross-entropy loss function, and perform backpropagation based on this to update and optimize the model parameters.

[0016] Further, evaluating the soothing effect of the cosmetic to be tested according to the number of fluorescent particles in each group includes: Taking the model control group as the standard, compare the number of fluorescent particles in each group with that of other groups. If there are significant differences between the data, the experimental data is valid; When the experimental data is valid, calculate the soothing effect of the cosmetic to be tested by using the number of fluorescent particles in the model control group, the normal control group and the group of samples to be tested; The second object of the present invention can be achieved by adopting the following technical solutions: A cosmetic soothing efficacy evaluation system based on a deep learning model, the system includes: An image acquisition module, used to set up a control group and a group of samples to be tested. After each group is cultured for the same time, methyl cellulose is used to fix zebrafish, and a side view of each group of zebrafish is taken under a fluorescence microscope as the original zebrafish image; among them, the control group includes a normal control group, a model control group and a positive control group. The normal control group is to put zebrafish into standard dilution water, the model control group is to put zebrafish into standard dilution water containing a modeling agent, and the positive control group is to put zebrafish into standard dilution water containing a modeling agent and dipotassium glycyrrhizinate; the group of samples to be tested is to put zebrafish into standard dilution water containing a modeling agent and a stock solution of the sample to be tested, and the stock solution of the sample to be tested is a solution prepared from the cosmetic to be tested; the zebrafish in each group are 2dpf neutrophil gene mutant zebrafish with normal development; An image preprocessing module, used to preprocess the original zebrafish image to improve the overall brightness of the image; A training dataset acquisition module, used to obtain a binary mask according to the preprocessed original zebrafish image; use the preprocessed original zebrafish image and the corresponding binary mask as samples in the training dataset; A model training module, used to train a semantic segmentation model by using the training dataset; A module for evaluating the cosmetic to be tested, used to input the preprocessed zebrafish images to be detected in each group into the trained semantic segmentation model to obtain images containing only the target area in each group; extract the number of fluorescent particles in each group according to the images containing only the target area in each group; evaluate the soothing effect of the cosmetic to be tested according to the number of fluorescent particles in each group.

[0017] The third object of the present invention can be achieved by adopting the following technical solutions: A terminal device, including a processor and a memory for storing programs executable by the processor. When the processor executes the programs stored in the memory, the above-mentioned cosmetic soothing efficacy evaluation method based on a deep learning model is realized.

[0018] The fourth object of the present invention can be achieved by adopting the following technical solutions: A computer-readable storage medium stores a program which, when executed by a processor, implements the above-mentioned method for evaluating the soothing effect of cosmetics based on a deep learning model.

[0019] The fifth object of the present invention can be achieved by adopting the following technical solutions: A computer program product includes a computer program which, when executed by a processor, implements the above-mentioned method for evaluating the soothing effect of cosmetics.

[0020] The present invention has the following beneficial effects compared with the prior art: The present invention combines fluorescence imaging technology and a deep learning model. The original image data of zebrafish is obtained by using zebrafish fluorescence imaging technology. The original image of zebrafish conforms to the data of the real reaction of the organism, which ensures the reliability of the obtained data. The semantic segmentation model is trained by using the preprocessed original image data of zebrafish. The trained semantic segmentation model can accurately segment the target area in the original image of zebrafish, and then automatically count the neutrophils in the target area, so as to realize the evaluation of the soothing effect of cosmetics. By combining fluorescence imaging technology and a deep learning model, not only the evaluation process is accelerated, but also the accuracy and scientificity of the evaluation of the soothing effect of cosmetics are greatly improved, providing new ideas for the development of the beauty industry. Description of the Drawings

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the structures shown in these drawings.

[0022] Figure 1 It is a flowchart of the method for evaluating the soothing effect of cosmetics based on a deep learning model in Embodiment 1 of the present invention; Figure 2 It is a schematic diagram of the principle of the method for evaluating the soothing effect of cosmetics based on a deep learning model in Embodiment 1 of the present invention; Figure 3 It is a schematic flow diagram of obtaining the original image of fluorescently labeled zebrafish in Embodiment 1 of the present invention; Figure 4 It is the original image of zebrafish in Embodiment 1 of the present invention; Figure 5 It is the preprocessed original image of zebrafish in Embodiment 1 of the present invention; Figure 6Among them, (a), (b), and (c) are respectively multiple pre-processed original images of zebrafish in Embodiment 1 of the present invention, a binary mask corresponding to the image in (a), and an image containing only the target area corresponding to the binary mask in (b); Figure 7 Among them, (a) and (b) are respectively multiple images containing only the target area in Embodiment 1 of the present invention, and the number of fluorescent particles extracted corresponding to the images in (a); Figure 8 is a structural block diagram of a cosmetic soothing efficacy evaluation system based on a deep learning model in Embodiment 2 of the present invention; Figure 9 is a structural block diagram of a terminal device in Embodiment 3 of the present invention. Detailed implementation manners

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. It should be understood that the specific embodiments described are only used to explain the present application and are not used to limit the present application.

[0024] Embodiment 1: As Figure 1 , 2 shown, this embodiment provides a method for evaluating the soothing efficacy of cosmetics based on a deep learning model, including the following steps: S101. Obtain the original image of zebrafish labeled with fluorescence.

[0025] As Figure 3 shown, the original image of zebrafish labeled with fluorescence is obtained through experiments, and the steps are as follows: (1) Confirmation of cosmetic samples. Confirm information such as the usage method, solubility, and stability of the cosmetic samples to be tested.

[0026] (2) Preparation of standard dilution water.

[0027] In one embodiment, 17.2 g of sodium chloride, 0.76 g of potassium chloride, 2.9 g of calcium chloride, and 4.9 g of magnesium sulfate are accurately weighed, dissolved with distilled water or deionized water and made up to 1000 ml. 16.67 ml is taken therefrom and diluted to 1000 ml with distilled water or deionized water to obtain the standard dilution water. Before use, it is balanced to oxygen saturation at (28 ± 2) °C with air and stored for later use.

[0028] (3)Preparation of the test cosmetic sample stock solution. The test cosmetic sample is formulated into a test sample stock solution at a certain concentration and stored for future use according to the characteristics of the test sample (at room temperature or refrigerated) as required.

[0029] (4)Sodium lauryl sulfate (SLS) is used as the modeling agent.

[0030] (5)Positive control. Dipotassium glycyrrhizinate is used as the positive control sample.

[0031] (6)Preliminary experiment.

[0032] (6-1)Select 2dpf neutrophil gene mutant zebrafish with normal development.

[0033] (6-2)The experimental groups are divided into a normal control group, a model control group, a positive control group, and a test sample group.

[0034] Normal control group: The zebrafish are placed in standard dilution water.

[0035] Model control group: The zebrafish are placed in standard dilution water containing SLS.

[0036] Test sample group: The zebrafish are placed in standard dilution water containing SLS and the test sample stock solution.

[0037] Positive control group: The zebrafish are placed in standard dilution water containing SLS and dipotassium glycyrrhizinate.

[0038] Each control group incubates zebrafish in a six-well plate, with 15 zebrafish incubated in each well.

[0039] (6-3)Wrap the culture vessels with aluminum foil and incubate at (28 ± 2) °C for 18 hours.

[0040] (6-4)Observe and record the death and other toxic effects of the zebrafish. The highest concentration at which no developmental abnormalities, deformities, death, and other toxic effects are observed in the zebrafish is used as the maximum detection concentration of the test sample.

[0041] (7)Formal experiment.

[0042] (7-1)The experimental grouping is the same as in the preliminary experiment. Dilute three concentrations in the test sample group by a factor of 2, such as 0.5%, 0.25%, 0.125%, and the highest test concentration shall not be higher than the maximum detection concentration.

[0043] (7-2)Repeat the operations in the preliminary experiment.

[0044] (7-3) Fix the zebrafish using methylcellulose and take pictures under a fluorescence microscope. Take the side view of the zebrafish for image analysis. First, select about 500 images from the images taken in each experimental group to train the segmentation model so that the model can learn the ability to segment the region of interest. After training the model, for any new cosmetic sample to be tested, only 10 images need to be selected from each experimental group as the valid experimental data for the current experimental group, and a total of 40 images from the 10 images of each experimental group are used as the original zebrafish images.

[0045] S102. Preprocess the original zebrafish images.

[0046] The fluorescence-labeled zebrafish images taken by the microscope have low brightness and are overall dark, making it impossible to clearly see the outline of the zebrafish. In order to clearly show the distinguishable yolk sac and the corresponding dorsal part above the lateral line, this embodiment preprocesses the original zebrafish images.

[0047] Brightness enhancement helps to improve the contrast between the foreground and the background, making it easier to identify the region of interest (ROI). By preprocessing the brightness of the image, the image can be made clearer and easier to view, and certain image noises can be eliminated, thereby improving the quality of the image, and it can provide a better input for subsequent image segmentation tasks.

[0048] The brightness processing in this embodiment converts the image from the RGB (Red, Green, Blue) color space to the HSV (Hue, Saturation, Value) color space, and the HSV color space separates the color information from the brightness information. By increasing the value of the V channel of the image, the overall brightness can be significantly improved, and the visibility of the image can be enhanced. Specifically, extract the V channel matrix of the HSV format image and multiply it element by element with the constant 1.3 to increase the value of the V channel matrix. The S channel (saturation channel) controls the intensity or purity of the color in the image. By increasing the value of the S channel in the same way, the color saturation can be enhanced, making the image more vivid. Then convert the original zebrafish image after brightness processing from HSV to the RGB space. From Figure 4 and Figure 5 It can be seen that after brightness processing, the image can clearly distinguish the zebrafish from the black background, while clearly identifying the fluorescent cells in the zebrafish body and clearly identifying the edges of the abdominal and dorsal regions of the zebrafish, facilitating subsequent processing and model training.

[0049] S103. Obtain the training data set according to the preprocessed original zebrafish images.

[0050] To accurately measure the soothing effect of cosmetics, it is necessary to observe the distribution and quantity of transgenic neutrophils in zebrafish. The fluorescent particles in the corresponding part of the zebrafish yolk sac and the back above the lateral line are important for subsequent analysis. Therefore, correctly segmenting the yolk sac and the corresponding part of the back above the lateral line (the target area) in each zebrafish image is crucial for accurately counting the particles. The target area in the image can be accurately outlined through a binary mask.

[0051] To obtain the binary mask corresponding to each image, engineers circle the target area in the image with a thin red line to obtain a preliminary line contour. Then, morphological operations such as erosion and dilation are used to further refine the line contour, removing possible noises or artifacts. These operations will help smooth the edges and fill any small gaps in the line contour. Finally, the line contour is filled with color and binarized to obtain the binary mask. The binary mask is an important part of the semantic segmentation model, which helps classify the pixels in the image as foreground or background. The binary mask is used as the label for the corresponding image.

[0052] A total of about 500 original zebrafish images are obtained from each experimental group. After preprocessing, the preprocessed original zebrafish images and the labels are used as samples in the training dataset.

[0053] S104. Use the training dataset to train the semantic segmentation model.

[0054] In the training stage, the preprocessed original zebrafish images are used as the input data for the semantic segmentation model. The model extracts and learns the features of the input data and outputs a binary mask. This binary mask is designed to accurately outline the yolk sac and the corresponding part of the back above the lateral line in the zebrafish image. Subsequently, the binary mask output by the model is compared with the binary mask corresponding to the preprocessed original zebrafish image. The difference value between the two is calculated using the cross-entropy loss function, and based on this, the model parameters are updated and optimized through the backpropagation algorithm. This enables the model to gradually improve the accuracy of the output binary mask during continuous iterative training, making it able to more accurately represent the real situation and achieve a more precise definition of the target area.

[0055] S105. Input the preprocessed zebrafish image to be detected into the trained semantic segmentation model to obtain an image containing only the target area.

[0056] First, perform image preprocessing operations on the original zebrafish images to be detected, and then input them into a semantic segmentation model that has been trained and has precise segmentation capabilities. The model outputs a binary mask (target region mask), which focuses on key target regions such as the yolk sac and the corresponding dorsal part above the lateral line in the zebrafish image. Perform a pixel-level multiplication operation on the target region mask and the original image to finally obtain a zebrafish image that only contains the target region.

[0057] This step combines technologies such as image preprocessing, deep learning models, and pixel-level masks to achieve efficient and accurate target detection tasks.

[0058] For the process from the preprocessed original zebrafish image to the image that only contains the target region, reference can be made to Figure 6 。

[0059] S106. Extract the number of fluorescent particles based on the image that only contains the target region; evaluate the soothing effect of the cosmetic according to the number of fluorescent particles.

[0060] In this step, the number of fluorescent particles is automatically extracted based on the image that only contains the target region; according to the number of fluorescent particles, the soothing effect of the cosmetic can be quickly and accurately evaluated.

[0061] (1) Extract the number of fluorescent particles based on the image that only contains the target region.

[0062] Since the fluorescent image is mainly composed of green pixel points, first extract the green channel from the RGB channels of the image and convert it into grayscale; then screen out the fluorescent particles corresponding to the pixels with pixel intensity greater than the set threshold, and then use the SimpleITK toolkit to set the threshold, and regard the connected components of the high-brightness pixel clusters as a single fluorescent particle, so as to obtain the number of fluorescent particles in the image. Reference can be made to Figure 7 。

[0063] (2) Evaluate the soothing effect of the cosmetic according to the number of fluorescent particles.

[0064] In this embodiment, two cosmetic samples to be tested are selected from the group of samples to be tested: Sample 16 and Sample 15. Randomly select 10 images from the zebrafish fluorescent images taken in each group as the experimental valid data. The number of fluorescent particles N extracted from each group is shown in Table 1.

[0065] Table 1 Number of fluorescent particles N extracted from each group

[0066] Step 1: Calculate the average value (Mean) and standard error (SE) of the neutrophil count for each group Normal control group: Data X iIt is: 11, 13, 13, 17, 17, 17, 18, 20, 21, 21 Average value calculation:

[0067] Standard error calculation: First calculate the variance , and the variance calculation formula is , where n is the number of data points.

[0068]

[0069]

[0070] Model control group: Data X i It is: 31, 32, 36, 37, 39, 39, 40, 41, 43, 44 Average value calculation:

[0071] Standard error calculation:

[0072] Sample 16: Data X i It is: 21, 23, 24, 25, 26, 26, 26, 29, 29, 30 Average value calculation:

[0073] Standard error calculation:

[0074] Sample 15: Data X i It is: 19, 23, 23, 24, 24, 25, 26, 28, 28, 28 Average value calculation:

[0075] Standard error calculation:

[0076] Positive control group: Data X i It is: 15, 15, 16, 16, 19, 20, 22, 22, 26, 28 Average value calculation:

[0077] Standard error calculation:

[0078] Step 2: Experimental validity conditions There was a statistically significant difference in the number of neutrophils between the normal control group and the model control group, and the difference in the mean values was greater than twice the standard deviation (SD) within the normal control group, demonstrating the effectiveness of this experiment.

[0079] There was a statistically significant difference in the number of neutrophils between the positive control group and the model control group, and the difference in the mean values was greater than twice the standard deviation (SD) within the model control group, demonstrating the effectiveness of this experiment.

[0080] Step 3: Conduct statistical analysis Based on the effectiveness of the experiment, the following statistical analysis was carried out: Check whether the data of each group conforms to a normal distribution. Using the statistical analysis function of SPSS, set the confidence interval to 95%.

[0081] Table 2 Detection results of whether the data of each group conforms to a normal distribution

[0082] When the sample size 3 ≤ n ≤ 5000, the results are based on the Shapiro-Wilk (W test). In the normality test results, since the sample size is 10, the results As are shown in Table 2. In the case of a 95% confidence interval (α = 0.05), P > 0.05, the null hypothesis is not rejected, so it is considered that the variable follows a normal distribution. Next, the t-test method was used to compare with the model control group.

[0083] Calculate the t-values of the model control group and other groups. The formula is:

[0084] where 、 are the mean values of the model control group and other groups, n 1 、n 2 are the number of data in the two groups, and s p is the combined standard deviation. The formula for s p is:

[0085] where 、 are the variances of the two groups.

[0086] Taking the test sample 16 as an example, first calculate the combined standard deviation of the model control group and the test sample 16:

[0087] Then calculate the t-value:

[0088] Next, calculate the degrees of freedom:

[0089] Finally, look up the t-distribution table. When the degrees of freedom d f = 18, look up the t-distribution table to find the p-value corresponding to t = 7.55 < 0.05.

[0090] The statistical analysis of the data of test sample 15 is similar.

[0091] The calculated p-values between the data of the two experimental groups of test sample 16 and test sample 15 and the data of the model control group are both < 0.05, which proves that there are significant differences between the data of the test sample group and the model control group. That is to say, the cosmetic sample is effective.

[0092] Step 4: Calculation of soothing effect When the cosmetic sample is effective, calculate the soothing effect of the test sample:

[0093] The calculated soothing effects of test sample 16 and test sample 15 are 57.48% and 62.62% respectively.

[0094] Those skilled in the art can understand that all or part of the steps in the method for implementing the above embodiments can be completed by instructing relevant hardware through a program, and the corresponding program can be stored in a computer-readable storage medium.

[0095] It should be noted that although the method operations of the above embodiments are described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the described steps can be changed in the execution order. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution.

[0096] Embodiment 2: As Figure 8 shown, this embodiment provides a cosmetic soothing effect evaluation system based on a deep learning model. The system includes an image acquisition module 801, an image preprocessing module 802, a training data set acquisition module 803, a model training module 804, and a test cosmetic evaluation module 805, where: The image acquisition module 801 is used to set up a control group and a test sample group. After each group is cultured for the same period of time, methylcellulose is used to fix zebrafish, and a lateral view of each group of zebrafish is taken under a fluorescence microscope as the original zebrafish image. Among them, the control group includes a normal control group, a model control group, and a positive control group. The normal control group is to place zebrafish in standard dilution water, the model control group is to place zebrafish in standard dilution water containing a modeling agent, and the positive control group is to place zebrafish in standard dilution water containing a modeling agent and dipotassium glycyrrhizinate. The test sample group is to place zebrafish in standard dilution water containing a modeling agent and a test sample stock solution, and the test sample stock solution is a solution prepared from the test cosmetic. The zebrafish in each group are 2dpf neutrophil gene mutant zebrafish with normal development. The image preprocessing module 802 is used to preprocess the original zebrafish image to improve the overall brightness of the image. The training dataset acquisition module 803 is used to obtain a binary mask based on the preprocessed original zebrafish image, and use the preprocessed original zebrafish image and the corresponding binary mask as samples in the training dataset. The model training module 804 is used to train a semantic segmentation model using the training dataset. The test cosmetic evaluation module 805 is used to input the preprocessed zebrafish images to be detected in each group into the trained semantic segmentation model to obtain images containing only the target area in each group. According to the images containing only the target area in each group, the number of fluorescent particles in each group is extracted. According to the number of fluorescent particles in each group, the soothing effect of the test cosmetic is evaluated.

[0097] For the specific implementation of each module in this embodiment, reference can be made to the above-mentioned Embodiment 1, which will not be elaborated here one by one. It should be noted that the system provided in this embodiment is only illustrated by the above division of each functional module. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure can be divided into different functional modules to complete all or part of the functions described above.

[0098] Embodiment 3: This embodiment provides a terminal device, which can be a computer, such as Figure 9As shown, it includes a processor 902, a memory, an input device 903, a display 904, and a network interface 905 connected via a system bus 901. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium 906 and an internal memory 907. The non-volatile storage medium 906 stores an operating system, a computer program, and a database. The internal memory 907 provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the processor 902 executes the computer program stored in the memory, it implements the method for evaluating the soothing efficacy of cosmetics based on a deep learning model in the above-mentioned Embodiment 1.

[0099] Embodiment 4: This embodiment provides a computer-readable storage medium that stores a computer program. When the computer program is executed by a processor, it implements the method for evaluating the soothing efficacy of cosmetics based on a deep learning model in the above-mentioned Embodiment 1.

[0100] It should be noted that the computer-readable storage medium in this embodiment can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0101] Embodiment 5: This embodiment provides a computer program product that includes a computer program. When the computer program is executed by a processor, it implements the method for evaluating the soothing efficacy of cosmetics in the above-mentioned Embodiment 1.

[0102] As mentioned above, only the preferred embodiments of this invention patent are described, but the protection scope of this invention patent is not limited thereto. Any person skilled in the art within the scope disclosed by this invention patent, according to the technical solution and inventive concept of this invention patent, makes equivalent substitutions or changes, and all belong to the protection scope of this invention patent.

Claims

1. A method for evaluating the soothing effect of cosmetics based on a deep learning model, characterized in that: The method comprises: Set up a normal control group: Place zebrafish in standard dilution water; Set up the model control group: put zebrafish into standard diluted water containing modeling agent; Setting up the sample group to be tested: placing the zebrafish into standard dilution water containing a modeling agent and a stock solution of the sample to be tested; the stock solution of the sample to be tested is a solution prepared from the cosmetics to be tested; Set up a positive control group: Place zebrafish in standard diluted water containing modeling agent and dipotassium glycyrrhizinate; The zebrafish in each group were 2 dpf neutrophil gene mutant zebrafish with normal development; After culturing for the same period of time, each group of zebrafish was fixed with methylcellulose, and the side view of each group of zebrafish was photographed under a fluorescence microscope as the original image of the zebrafish; Preprocessing the original zebrafish image to improve the overall brightness of the image; According to the preprocessed zebrafish original image, a binary mask is obtained; the preprocessed zebrafish original image and the corresponding binary mask are used as samples in a training data set; Use the training dataset to train the semantic segmentation model; The zebrafish images to be tested in each group are preprocessed and input into the trained semantic segmentation model to obtain images of each group containing only the target area; based on the images of each group containing only the target area, the number of fluorescent particles in each group is extracted; based on the number of fluorescent particles in each group, the soothing effect of the cosmetics to be tested is evaluated.

2. The method for evaluating the soothing effect of cosmetics according to claim 1, characterized in that: The zebrafish images to be detected in each group are pre-processed and then input into the trained semantic segmentation model to obtain images in each group that only contain the target area, including: The zebrafish images to be detected in each group are preprocessed and input into the trained semantic segmentation model, and the corresponding binary masks are output; The binary mask is multiplied by the corresponding preprocessed zebrafish image to be detected at the pixel level to obtain a zebrafish image containing only the target area.

3. The method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 2, characterized in that: The preprocessing comprises: Convert the image from RGB space to HSV space; Increase the pixel values ​​of the V channel and S channel of the image to obtain an image with increased brightness; Convert the image with increased brightness from HSV space to RGB space; The image is an original zebrafish image or an image of a zebrafish to be detected in each group.

4. The method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 2, characterized in that: The binary mask is obtained according to the preprocessed zebrafish original image, including: The target area is circled in the preprocessed original image of the zebrafish with a thin line to obtain a preliminary line outline; the target area is the zebrafish yolk sac and two areas corresponding to the back above the lateral line; Based on the preliminary line contour, morphological operations are used to refine the line contour and remove the existing noise or artifacts; Finally, the line contours are color-filled and binarized to obtain a binary mask.

5. The method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 2, characterized in that: The step of extracting the number of fluorescent particles in each group according to the image of each group containing only the target area includes: The green channel is extracted from the RGB channels of the image containing only the target area and converted into a grayscale image. The pixels in the grayscale image whose values ​​are greater than the set threshold are filtered out, and the connected pixels are regarded as a fluorescent particle using the SimpleITK toolkit to obtain the number of fluorescent particles in the image.

6. The method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 2, characterized in that: The method of training the semantic segmentation model using the training data set includes: The preprocessed zebrafish raw image is input into the semantic segmentation model and the binary mask is output; The output binary mask is compared with the binary mask corresponding to the preprocessed original zebrafish image, and the difference between the two is calculated using the cross entropy loss function, which is then used as a basis for backpropagation to update the optimization model parameters.

7. The method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 2, characterized in that: The soothing effect of the cosmetics to be tested is evaluated according to the number of fluorescent particles in each group, including: Take the model control group as the standard and compare the number of fluorescent particles in other groups. If there is a significant difference between the data, the experimental data is valid. When the experimental data is valid, the soothing effect of the cosmetics to be tested is calculated using the number of fluorescent particles in the model control group, normal control group and test sample group.

8. A cosmetic soothing efficacy evaluation system based on a deep learning model, characterized in that: The system comprises: The image acquisition module is used to set a control group and a sample group to be tested. After culturing for the same time, each group uses methylcellulose to fix the zebrafish, and takes a side view of each group of zebrafish under a fluorescence microscope as the original image of the zebrafish; wherein the control group includes a normal control group, a model control group and a positive control group, the normal control group is to put the zebrafish in standard dilution water, the model control group is to put the zebrafish in standard dilution water containing a modeling agent, and the positive control group is to put the zebrafish in standard dilution water containing a modeling agent and dipotassium glycyrrhizinate; the sample group to be tested is to put the zebrafish in standard dilution water containing a modeling agent and a sample stock solution to be tested, and the sample stock solution to be tested is a solution prepared by the cosmetics to be tested; the zebrafish in each group are 2dpf neutrophil gene mutant zebrafish with normal development; An image preprocessing module, used for preprocessing the original zebrafish image to improve the overall brightness of the image; The training data set acquisition module is used to obtain a binary mask based on the preprocessed zebrafish original image; the preprocessed zebrafish original image and the corresponding binary mask are used as samples in the training data set; A model training module is used to train the semantic segmentation model using the training dataset; The cosmetics to be tested evaluation module is used to input the zebrafish images to be tested in each group into the trained semantic segmentation model after preprocessing, so as to obtain images of each group containing only the target area; based on the images of each group containing only the target area, the number of fluorescent particles in each group is extracted; and based on the number of fluorescent particles in each group, the soothing effect of the cosmetics to be tested is evaluated.

9. A terminal device, comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, the method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for evaluating the soothing effect of cosmetics according to any one of claims 1 to 7 is implemented.

Images