KR102167185B1 - Skin test method and cosmetic composition manufacturing method using the same - Google Patents

Abstract

Disclosed are a skin test method and a cosmetic composition manufacturing method using the same. According to one embodiment of the present invention, the cosmetic composition manufacturing method based on artificial intelligence comprises: separately obtaining a first facial image and a second facial image by controlling a camera in a skin analyzer under a first light source and a second light source of the skin analyzer; generating a first image vector on the basis of the first facial image and the second facial image; obtaining a first image output by applying the first image vector to a pre-learned first convolutional neural network; obtaining a skin range on the basis of the first image output, and obtaining color information and gloss information within the skin range on the basis of the skin range; generating a second image vector on the basis of the skin range, and applying the second image vector to pre-learned second and third convolutional neural networks so as to obtain second and third image outputs, respectively; and obtaining first skin quality information on the basis of the second image output, and obtaining second skin quality information on the basis of the third image output. According to the present invention, the skin range for a skin test is determined.

Description

Skin test method and cosmetic composition manufacturing method using the same {SKIN TEST METHOD AND COSMETIC COMPOSITION MANUFACTURING METHOD USING THE SAME}

The following examples relate to a skin test method and a method for preparing a cosmetic composition using the same.

Cosmetic compositions for skin beauty may have different functions depending on their raw material ingredients. In particular, use of a cosmetic composition that is not suitable for the skin may cause excessive irritation to the skin. Therefore, the use of a customized cosmetic composition is very important, and for this, it is necessary to systematize a skin test method, and to optimize the type and content ratio of raw materials. However, despite the existence of various skin testing methods, the design of a system integrating them is often insufficient, and there is no system to systematize the types and contents of raw materials. Therefore, there is a need for a study on a skin test method using artificial intelligence and a method for selecting a raw material and adjusting its content ratio based on it, and testing it virtually.

KR101792541 KR101444726 KR102018446 KR101329146

Examples are to determine the skin range for skin testing by applying deep learning technology.

Examples intend to evaluate the condition of skin quality by applying deep learning technology.

Examples are intended to determine the type and content ratio of optimized ingredients suitable for individual skin by applying deep learning technology.

Examples are to predict the result of skin application from the condition of the skin and the composition of the cosmetic composition by applying deep learning technology.

The embodiments aim to convert information into big data through a blockchain network and use enhanced security.

In the method of manufacturing a cosmetic composition based on artificial intelligence according to an embodiment, the first and second face images are obtained by controlling a camera in the skin analyzer under a first light source and a second light source of the skin analyzer. Step to do; Generating a first image vector based on the first face image and the second face image; Applying the first image vector to a pre-trained first convolutional neural network to obtain a first image output; Skin range based on the output of the first image-the skin range includes the first face image and the second face image, and the range of the forehead and cheek recognized as skin in the first and second face images Means-obtaining; Obtaining color information and gloss information within the skin range, based on the skin range; Generating a second image vector based on the skin range; Applying the second image vector to a pre-trained second convolutional neural network and a third convolutional neural network to obtain a second image output and a third image output, respectively; Based on the output of the second image, the first skin quality information-the first skin quality information includes structures to be improved within a range of skin that can be primarily distinguished from the face, and the structures to be improved include acne, wrinkles, gombo, and atopy. Dermatitis, erythema, urticaria and eczematous skin disease means-obtaining; And acquiring second skin quality information based on the output of the third image, wherein the second skin quality information means information on particles of an epidermis capable of expressing an overall management state of the face, and the first The light source is yellow light in a wavelength range of 576 to 580 nm, the second light source is cyan light in a wavelength range of 490 to 520 nm, and the color information is graphic information expressed in RGB colors of 8-bit 256 colors for each pixel in the skin range. The glossiness information refers to a numerical value representing the intensity of light when the reflected light is received at the same angle as the incident light on the surface of the skin area, and the first convolutional neural network is a labeled training first face image Obtaining second face images and labeled training; Generating first training image vectors based on the labeled training first face images and the labeled training second face images; Applying the first training image vectors to the first convolutional neural network to obtain first training image outputs; And training the first convolutional neural network based on the first training image outputs and the first image labels, wherein the second convolutional neural network and the third convolutional neural network have a labeled training skin range. Obtaining them; Based on the labeled training skin ranges, generating second training image vectors; Applying the second training image vectors to the second convolutional neural network and the third convolutional neural network to obtain second training image outputs and third training image outputs; And training the second convolutional neural network and the third convolutional neural network based on the second training image outputs, the third training image outputs, second image labels, and third image labels. can do.

According to an embodiment, the first to the first skin regions, each of which is rubbed for 30 seconds, on the face regions corresponding to the first skin region, the second skin region, and the third skin region, each of which is subdivided into the skin region, is applied to the test kit. 3 obtaining epidermal content information-the epidermal content information includes the content and content ratio of oil, moisture, and contaminants that can be obtained from the epidermis; Generating a first sensing vector based on the first to third epidermal content information; Obtaining a first sensing output by applying the first sensing vector to a first neural network; And obtaining integrated epidermal containing information based on the first sensing output, wherein the first neural network is based on facial regions corresponding to first to third skin regions of the skin ranges, Acquiring labeled training first to third epidermal content information; Generating first training sensing vectors based on the labeled training first to third epidermal containing information; Applying the first training sensing vectors to the first neural network to obtain first training sensing outputs; And training the first neural network based on the first training sensing outputs and first sensing labels.

According to an embodiment, the step of preparing a raw material based on data obtained through the skin test method of claim 1 to claim 2; Mixing the prepared raw materials to prepare a cosmetic composition; And it may include the step of packaging the prepared cosmetic composition.

According to an embodiment, in the preparing of the raw material, based on the color information, the gloss information, the first skin quality information, the second skin quality information, and the integrated epidermal content information, the first integrated vector is Generating; Applying the first integrated vector to a second neural network to obtain a first integrated output; Obtaining a raw material composition ratio based on the first integrated output; Generating a second integrated vector based on the first integrated vector and the raw material composition ratio; Applying the second unified vector to a third neural network to obtain a second unified output; Based on the second integrated output, step-by-step predicted data-the step-by-step predicted data predicts the change in skin according to the amount of one application of the cosmetic composition, the number of applications, and the application period when applying the cosmetic composition generated based on the raw material composition ratio Obtaining the analyzed data; Obtaining feedback information based on the step-by-step predicted data; Correcting the raw material composition ratio based on the feedback information; Preparing the raw material based on the modified raw material composition ratio; And storing the second integrated vector and the feedback information, wherein the second neural network includes labeled training color information, labeled training gloss information, labeled training first skin quality information, and a label Obtaining de-training second skin quality information and labeled training integrated epidermal content information; Based on the labeled training color information, the labeled training gloss information, the labeled training first skin quality information, the labeled training second skin quality information, and the labeled training integrated epidermal content information Generating first training integration vectors; Applying the first training integration vectors to the second neural network to obtain first training integration outputs; And training the second neural network based on the first unified training outputs and the first unified labels, wherein the third neural network includes a raw material composition ratio based on the first unified training outputs. Obtaining them; Generating second training integration vectors based on the first training integration vectors and the raw material composition ratios; Applying the second training integration vectors to the third neural network to obtain second training integration outputs; And training the third neural network based on the second training unified outputs and second unified labels.

According to an embodiment, the storing of the second integration vector and the feedback information includes storing the second integration vector and the feedback information in a blockchain network, and the blockchain network includes the second integration Blocks including a vector and the feedback information; Chains connecting the blocks in chronological order; And a private blockchain network including the network storage devices for storing the blocks and the chains, wherein the network storage devices include: a first network storage device including a raw material manufacturer of the cosmetic composition; A second network storage device including a manufacturer that manufactures the cosmetic composition; A third network storage device including a distributor of the cosmetic composition; And an Internet connection network connecting each of the network storage devices.

The apparatus according to the embodiment may be controlled by a computer program stored in a medium in order to execute the method of any one of the above-described methods in combination with hardware.

The embodiments may determine a skin range for testing the skin by applying deep learning technology.

Examples can evaluate the condition of skin quality by applying deep learning technology.

Examples may determine the type and content ratio of optimized ingredients suitable for individual skin by applying deep learning technology.

Examples may predict the result of skin application from the condition of the skin and the composition of the cosmetic composition by applying deep learning technology.

The embodiments can convert information into big data and use enhanced security through a blockchain network.

1 is a flowchart illustrating a method of testing skin based on facial images according to an exemplary embodiment.
2 is a diagram for explaining convolutional neural networks used to process face images according to an embodiment.
FIG. 3 is a diagram for describing a learning method of a first convolutional neural network used to process face images according to an exemplary embodiment.
FIG. 4 is a diagram for describing a learning method of second to third convolutional neural networks used to process a skin range according to an exemplary embodiment.
5 is a diagram illustrating a neural network employed to process epidermal-containing information according to an embodiment.
6 is a diagram for explaining a learning method employed to process epidermal-containing information according to an exemplary embodiment.
7 is a diagram for explaining neural networks employed to obtain a raw material composition ratio according to an exemplary embodiment.
8 is a diagram for explaining a learning method employed to obtain a raw material composition ratio according to an embodiment.
9 is a diagram illustrating a block chain network according to an embodiment.
10 is an exemplary diagram of a configuration of an apparatus according to an embodiment according to an embodiment.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. However, since various changes may be made to the embodiments, the scope of the rights of the patent application is not limited or limited by these embodiments. It should be understood that all changes, equivalents, or substitutes to the embodiments are included in the scope of the rights.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only, and may be changed in various forms and implemented. Accordingly, the embodiments are not limited to a specific disclosure form, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical idea.

Although terms such as first or second may be used to describe various components, these terms should be interpreted only for the purpose of distinguishing one component from other components. For example, a first component may be referred to as a second component, and similarly, a second component may be referred to as a first component.

When a component is referred to as being "connected" to another component, it is to be understood that it may be directly connected or connected to the other component, but other components may exist in the middle.

The terms used in the examples are used for illustrative purposes only and should not be interpreted as limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present specification, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, but one or more other features. It is to be understood that the presence or addition of elements or numbers, steps, actions, components, parts, or combinations thereof, does not preclude in advance.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. Terms as defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in this application. Does not.

In addition, in the description with reference to the accompanying drawings, the same reference numerals are assigned to the same components regardless of the reference numerals, and redundant descriptions thereof will be omitted. In describing the embodiments, when it is determined that a detailed description of related known technologies may unnecessarily obscure the subject matter of the embodiments, the detailed description thereof will be omitted.

The embodiments may be implemented in various types of products such as a personal computer, a laptop computer, a tablet computer, a smart phone, a television, a smart home appliance, an intelligent vehicle, a kiosk, and a wearable device.

1 and 2 are flow charts and diagrams for explaining a method of testing skin based on facial images according to an exemplary embodiment.

According to an embodiment, a control device (hereinafter, a control device) used to manufacture a cosmetic composition based on artificial intelligence controls the camera in the skin analyzer under the first light source 201 and the second light source 202 of the skin analyzer. In addition, the first face image 203 and the second face image 204 may be obtained (101).

The control device according to an embodiment is a device that performs skin testing and manufacture of a cosmetic composition, and may be implemented as, for example, a hardware module, a software module, or a combination thereof.

According to an embodiment, the control device may be implemented as a server. The server may communicate with the terminal, retrieve necessary information from the database, employ it for a determination operation, and update the database based on the result of the determination. For example, the control device controls the user's camera to acquire images and photos in real time or through memory storage, updates face image-related information based on the acquired images and photos, and based on the updated information. The type and category of the automatic correction filter to be acquired can be updated.

According to an embodiment, the control device is a GPS (Global Positioning System) based positioning technology, a Wi-Fi based positioning technology, a mobile communication or cellular based positioning technology such as WCDMA, WiBro, 3G, LTE, 5G, WiMAX, etc. The user's location can be acquired through various methods such as communication (RFID, ZigBee, NFC, Bluetooth) and other multiple infrastructures, and various techniques can be applied and applied to the location acquisition method. The control device may be implemented with at least one of a server and a terminal. The operation performed by the control device may be implemented by either the server or the terminal alone, but may be implemented by the interaction of the server and the terminal. The server and terminal may include electronic devices.

According to an embodiment, the electronic device may be a smart home appliance having a communication function. Smart home appliances, for example, include televisions, digital video disk (DVD) players, audio, refrigerators, air conditioners, vacuum cleaners, ovens, microwave ovens, washing machines, air purifiers, set-top boxes, and TVs. It may include at least one of a box (eg, Samsung HomeSyncTM, Apple TVTM, or Google TVTM), game consoles, electronic dictionary, electronic key, camcorder, or electronic frame.

According to an embodiment, the electronic device includes various medical devices (e.g., magnetic resonance angiography (MRA), magnetic resonance imaging (MRI), computed tomography (CT)), an imager, an ultrasonic device, etc.), a navigation device, a GPS receiver ( global positioning system receiver), EDR (event data recorder), FDR (flight data recorder), automobile infotainment device, marine electronic equipment (e.g., marine navigation equipment and gyro compass, etc.), avionics, security It may include at least one of a device, a vehicle head unit, an industrial or domestic robot, an automatic teller's machine (ATM) of a financial institution, or a point of sales (POS) of a store.

According to one embodiment, the electronic device is a piece of furniture or building/structure including a communication function, an electronic board, an electronic signature receiving device, a projector, or various measurement devices. It may include at least one of devices (eg, water, electricity, gas, or radio wave measurement devices). The electronic device according to an embodiment may be a combination of one or more of the aforementioned various devices. In addition, the electronic device according to an embodiment may be a flexible device. In addition, it is obvious to those skilled in the art that the electronic device according to the embodiment is not limited to the above-described devices. The term user used in various embodiments may refer to a person using an electronic device or a device (eg, an artificial intelligence electronic device) using an electronic device.

An electronic device according to an embodiment includes a processor, a memory, a user interface, and a communication interface, and may be connected to other electronic devices through a network. The communication interface may transmit and receive data to and from other electronic devices within a predetermined distance through wired, wireless network, or wired serial communication. The network enables wired and wireless communication between an electronic device and various entities according to an embodiment. The electronic device can communicate with various entities through the network, and the network can use standard communication technologies and/or protocols. In this case, the network includes, but is not limited to, the Internet, a local area network (LAN), a wireless local area network (LAN), a wide area network (WAN), a personal area network (PAN), and the like, It can be understood by those of ordinary skill in the communication technology field that it may be another type of network capable of transmitting and receiving information.

The portable terminal according to an embodiment may be an electronic device including a communication function. For example, electronic devices include smart phones, tablet personal computers (PCs), mobile phones, video phones, e-book readers, desktop personal computers (desktop personal computers), and laptops. Laptop personal computer (PC), netbook computer, personal digital assistant (PDA), portable multimedia player (PMP), MP3 player, mobile medical device, camera, or wearable device (e.g.: Including at least one of a head-mounted-device (HMD) such as an electronic glasses, an electronic clothing, an electronic bracelet, an electronic necklace, an electronic appcessory, an electronic tattoo, a smart car, or a smartwatch. I can.

The first light source 201 according to an embodiment may be yellow light having a wavelength of 576 to 580 nm, and the second light source 202 may be cyan light having a wavelength of 490 to 520 nm. The use of light sources with different wavelengths can more accurately measure information according to the depth of the skin, and the condition of the skin can be more closely checked from the wavelengths reflected, refracted, and transmitted.

The type of camera according to an embodiment may be different according to the type and specification of the skin analyzer. In general, the camera may be a camera of FHD (Full High Definition) image quality, and may be equipment for capturing images and photos. The camera can automatically adjust ISO (International Organization for Standardization), exposure, focus, shutter speed, and zoom, but is not limited thereto. When shooting video with a camera, 60 frames are mainly used, and in some cases, more or less than that.

According to an embodiment, the control device may acquire face images including a face, and a corresponding first face image 203 and a corresponding first face image 203 from a photograph taken under the first light source 201 and the second light source 202, respectively. A second face image 204 may be obtained. The first face image 203 and the second face image 204 are photographed while the head is turned over so that the forehead is visible, and may be photographed while the obstructive foreign substances including glasses, bands, patches, and earrings are removed. Not limited.

According to an embodiment, the control device may generate the first image vector 210 based on the first face image 203 and the second face image 204 (102).

According to an embodiment, the control device may generate the first image vector 210 from the first face image 203 and the second face image 204, wherein the first image vector 210 is a first face image. 203 and the second face image 204 may be a multidimensional vector to which a weighted sum of RGB colors, saturation and brightness for each pixel is applied, and the weights applied to the weighted sum will be adjusted so that each feature increases as the distance from the average increases. I can. Here, the vector may be defined in various forms according to design intention, such as a one-hot vector and a real vector. When generating a feature vector, weights may be applied to various variables in the access request packet. Weights can be optimized during training of the convolutional neural network.

According to an embodiment, the control device may obtain the first image output 212 by applying the first image vector 210 to the pre-trained first convolutional neural network 211 (103).

According to an embodiment, the first image vector 210 may correspond to an input layer of the first convolutional neural network 211. According to an embodiment, the control device may obtain the first image output 212 generated from the first convolutional neural network 211 to which the first image vector 210 is applied. The first image output 212 may correspond to an output layer of the first convolutional neural network 211.

According to an embodiment, the control device may acquire the skin range 205 based on the first image output 212 (104 ). The skin range 205 includes a first face image 203 and a second face image 204, and a range of the forehead and cheeks recognized as skin in the first face image 203 and the second face image 204 Can mean In the cosmetic composition manufacturing method, the skin range 205 may be subdivided into a first skin area, a second area, and a third skin area, and the first skin area may include a T-shaped area descending from the forehead to the brow, The second skin region and the third skin region may be regions including left and right cheeks, respectively.

According to an embodiment, the control device may determine the skin range 205 for testing the skin by applying the deep learning technology.

According to an embodiment, the control device may acquire color information 206 and gloss information 207 within the skin range 205 based on the skin range 205 (105).

The color information 206 according to an embodiment may refer to graphic information expressed in RGB colors of 8-bit 256 colors for each pixel of the skin range 205. The expression of RGB colors can be achieved by adding and mixing RGB, and by allocating 8 bits to the corresponding red, green, and blue colors, 256 types of brightness and saturation can be determined for each color. Through this system, 16,777,216 colors, saturation, and brightness of 256 ³ can be expressed. For example, black color can be expressed as (0,0,0) in a three-dimensional Cartesian coordinate system of three primary colors in which red, green, and blue colors are numerically respectively, and white color can be expressed as (255,255,255). Red may be (255,0,0), green may be (0,255,0), and blue may be (0,0,255). In some cases, the color information 206 may use a CMYK subtraction and mixing method.

The glossiness according to an embodiment represents the intensity of light when the reflected light is received at the same angle as the incident light, and the glass surface gloss having a refractive index of 1.567 is 100, and the general angle may be 60°. Glossiness can be based on ISO 2813 and ASTM D523, and can generally be measured in GU units. Standard gloss can be measured at 60°, and in some cases, an angle of 20° (low gloss) to 85° (high gloss) can be used.

According to an embodiment, the control device may generate the second image vector 220 based on the skin range 205 (106 ).

According to an embodiment, the control device may generate the second image vector 220 from the skin range 205, wherein the second image vector 220 is RGB color and saturation for each pixel corresponding to the skin range 205 And a multidimensional vector to which the weighted sum of the luminances is applied, and weights applied to the weighted sum may be adjusted to increase as the features move away from the average. Here, the vector may be defined in various forms according to design intention, such as a one-hot vector and a real vector. When generating a feature vector, weights may be applied to various variables in the access request packet. Weights can be optimized during training of the convolutional neural network.

According to an embodiment, the control device applies the second image vector 220 to the pre-trained second convolutional neural network 221 and the third convolutional neural network 231 to output the second image 222 and the third Each of the image outputs 232 may be obtained (107).

According to an embodiment, the second image vector 220 may correspond to an input layer of the second convolutional neural network 221 and the third convolutional neural network 231. According to an embodiment, the control device includes a second image output 222 and a third image generated from the second convolutional neural network 221 and the third convolutional neural network 231 to which the second image vector 220 is applied. Output 232 can be obtained. The second image output 222 and the third image output 232 may respectively correspond to output layers of the second convolutional neural network 221 and the third convolutional neural network 231.

According to an embodiment, the control device may obtain the first skin quality information 208 based on the second image output 222 (108). The first skin quality information 208 may include structures to be improved within the skin range 205 that is primarily distinguishable from the face. Structures to be improved may include, but are not limited to, acne, wrinkles, gombo, atopic dermatitis, erythema, urticaria and eczematous skin diseases. Acne refers to clogged pores, pimples, and deep boils, and can appear as inflammatory skin diseases such as comedones, papules, pus blisters, nodules, and false sacs. Wrinkles may refer to a phenomenon in which skin moisture decreases due to degeneration of collagen fibers, elastic fibers, etc. due to various factors, resulting in decreased skin elasticity and skin folds. Gombo may refer to a skin condition in which scars on the face remain after suffering from infectious diseases such as smallpox or skin diseases such as acne. Atopic dermatitis is a chronic recurrent inflammatory skin disease accompanied by itchiness and dry skin, and may be characterized by red spots or rough skin. Erythema can be a form of reddening of the skin and dilated blood vessels caused by external and internal stimuli, resulting in blood accumulation. Hives can be swelling of various sizes caused by partial swelling of the upper layers of the skin. Eczema skin disease is a disease accompanied by itching. In the acute phase, blisters, erythema, and swelling appear, and in the chronic phase, the skin thickens and scales appear.

According to an embodiment, the control device may obtain the second skin quality information 209 based on the third image output 232 (109 ). The second skin quality information 209 may mean information on particles of an epidermis capable of expressing the overall management status of the face. The particle information is an index indicating the roughness of the skin, and may include a total of 4 steps of particle information according to the roughness. The first-stage particles are the smallest particles, and may mean that the skin has an excellent and smooth condition. The second-stage particles are small particles, and may mean a condition in which the skin is well maintained and the pore size is also appropriate. The third-stage particles are large particles, and may mean a condition in which skin care is poor and pores are large. The fourth-stage particles are very large particles, and may mean a condition in which skin care is very poor, and the size and number of pores are irregular and large.

According to an embodiment, the control device may evaluate the condition of skin quality by applying deep learning technology.

FIG. 3 is a diagram for explaining a learning method of a first convolutional neural network 304 used to process face images according to an exemplary embodiment.

According to an embodiment, the learning device may train a convolutional neural network for processing face images. The learning device may be a separate subject different from the control device, but is not limited thereto.

According to an embodiment, the artificial neural network includes an input layer to which training samples are input and an output layer to output training outputs, and may be trained based on a difference between the training outputs and labels. Here, the labels may be defined based on face images and identification information corresponding to the face images. The artificial neural network is connected by a group of a plurality of nodes, and is defined by weights between the connected nodes and an activation function that activates the nodes.

The learning device may train an artificial neural network using a gradient descent (GD) technique or a stochastic gradient descent (SGD) technique. The learning device can use the loss function designed by the outputs and labels of the artificial neural network.

The learning device may calculate a training error using a predefined loss function. The loss function can be predefined with a label, an output and a parameter as input variables, where the parameter can be set by weights in the artificial neural network. For example, the loss function may be designed in a mean square error (MSE) form, an entropy form, or the like, and various techniques or methods may be employed in an embodiment in which the loss function is designed.

The learning apparatus may find weights that influence the training error using a backpropagation technique. Here, the weights are relations between nodes in the artificial neural network. The learning apparatus can use the SGD technique using labels and outputs to optimize the weights found through the backpropagation technique. For example, the learning apparatus may update weights of the loss function defined based on labels, outputs, and weights using the SGD technique.

According to an embodiment, the learning apparatus may acquire labeled training first face images 301 and labeled training second face images 302. The training device may acquire training first face images, training second face images, and pre-labeled information, respectively, wherein the first face images and second face images are to be labeled according to positions of pre-classified skin regions. I can. The training data used at this time may be determined according to 10,000 pairs of labeled training first face images 301 and labeled training second face images 302 and skin ranges generated by skin experts. have.

According to an embodiment, the learning apparatus may generate first training image vectors 303 based on the labeled training first face images 301 and the labeled training second face images 302. .

According to an embodiment, the learning apparatus may obtain first training image outputs 305 by applying the first training image vectors 303 to the first convolutional neural network 304. The training device may train the first convolutional neural network 304 based on the first training image outputs 305 and the first image labels 306. The learning apparatus calculates training errors corresponding to the first training design outputs, and optimizes the connection relationship of nodes in the first convolutional neural network 304 to minimize the training errors, thereby constructing the first convolutional neural network 304. You can learn. The control device may acquire a skin range from the first face image and the second face image by using the learned first convolutional neural network 304.

4 is a diagram for describing a learning method of the second to third convolutional neural networks 420 used to process a skin range according to an exemplary embodiment.

According to an embodiment, the learning device may train a convolutional neural network required to acquire first skin quality information and second skin quality information from a skin range. The learning device may be a separate subject different from the control device, but is not limited thereto, and contents overlapping with the above-described learning method are omitted.

According to an embodiment, the learning device may acquire the labeled training skin ranges 401. The learning device may obtain pre-labeled information on each of the training skin ranges 401, and the training skin ranges 401 may be labeled according to pre-classified first skin quality information and second skin quality information. I can. The learning data used at this time may be determined according to 10,000 skin ranges and first skin quality information and second skin quality information generated by skin experts.

According to an embodiment, the learning device may generate second training image vectors 402 based on the labeled training skin ranges 401.

According to an embodiment, the learning apparatus applies the second training image vectors 402 to the second convolutional neural network 410 and the third convolutional neural network 420 to provide the second training image outputs 411 and the second training image vectors 402. 3 training image outputs 421 may be obtained. The training device is based on the second training image outputs 411, the third training image outputs 421, the second image labels 412, and the third image labels 422, based on the second convolutional neural network 410 ) And the third convolutional neural network 420 may be trained. The training apparatus calculates training errors corresponding to the second training image outputs 411 and the third training image outputs 421, and uses a second convolutional neural network 410 and a third convoluted network to minimize the training errors. The second convolutional neural network 410 and the third convolutional neural network 420 may be trained by optimizing a connection relationship between nodes in the lusion neural network 420. The control device may acquire the first skin quality information and the second skin quality information from the skin range using the second convolutional neural network 410 and the third convolutional neural network 420 on which the learning has been completed.

5 is a diagram illustrating a neural network employed to process epidermal-containing information according to an embodiment.

Referring to FIG. 5, the control device is for examination by rubbing each of the face regions corresponding to the first skin region 510, the second skin region 520, and the third skin region 530 into a subdivided skin region for 30 seconds. The first to third epidermal content information 511, 521, 531 may be obtained by applying the cotton to the test kit. The epidermal content information may include the content and content ratio of oil, moisture, and contaminants that can be obtained from the epidermis.

The subdivided areas of the skin range according to an embodiment are as described above. Cotton for examination may be made of the same material as a general cotton pad, but may be in a sterile mark to minimize contamination. When using inspection cotton, it may be better to rub it lightly than to rub it hard. The test kit is an equipment for checking the content of substances in the epidermis obtained through the test cotton, and can be confirmed through measurements such as polarization, density, electrical conductivity, and chromatography, but is not limited thereto.

According to an embodiment, the control device may generate the first sensing vector 501 based on the first to third skin-containing information 511, 521, and 531. The first sensing vector 501 may be a multidimensional vector to which a weighted sum of values of the first to third epidermal containing information 511, 521, 531 obtained from the first to third skin regions 530 is applied. The weights applied to the sum can be adjusted so that each feature increases as it moves away from the average. Here, the vector may be defined in various forms according to design intention, such as a one-hot vector and a real vector. When generating a feature vector, weights may be applied to various variables in the access request packet. Weights can be optimized during training of artificial neural networks.

According to an embodiment, the control device may apply the first sensing vector 501 to the pre-learned first neural network 502. The first sensing vector 501 may correspond to an input layer of the first neural network 502.

According to an embodiment, the control device may obtain the first sensing output 503 generated from the first neural network 502 to which the first sensing vector 501 is applied. The first sensing output 503 may correspond to an output layer of the first neural network 502. The control device may obtain the integrated skin content information 504 based on the first sensing output 503. The integrated epidermal content information 504 is not simply an average of the first to third epidermal content information 511, 521, 531, but can be expressed as a discrete or continuous value that is graded and expressed for each face area. have.

According to an embodiment, the control device may apply different skin characteristics according to gender, age, and race by querying the database from the identification information. The control device may train the neural network to obtain the integrated skin-containing information 504 from the first to third skin-containing information 511, 521, and 531. Contents related to learning for processing epidermal-containing information will be described later with reference to FIG. 6.

6 is a diagram for explaining a learning method employed to process epidermal-containing information according to an exemplary embodiment.

According to an embodiment, the learning device may learn a neural network required to acquire integrated epidermal content information from the first to third epidermal content information. The learning device may be a separate subject different from the control device, but is not limited thereto, and contents overlapping with the above-described learning method are omitted.

According to an embodiment, the learning device includes the first to third epidermal containing information of the labeled training based on the facial regions corresponding to the first to third skin regions 610, 620, and 630 of the skin ranges. 611, 621, 631) can be obtained. The learning device may acquire pre-labeled information on the training first to third epidermal-containing information 611, 621, and 631, respectively, and the training first to third epidermal-containing information 611, 621, and 631 It can be labeled according to pre-classified integrated epidermal content information. The learning data used at this time may be determined according to 10,000 pairs of first to third epidermal content information and integrated epidermal content information generated by skin experts.

According to an embodiment, the learning device may generate first training sensing vectors 601 based on the labeled training first to third epidermal-containing information 611, 621, and 631.

According to an embodiment, the learning apparatus may obtain first training sensing outputs 603 by applying the first training sensing vectors 601 to the first neural network 602. The learning apparatus may train the first neural network 602 based on the first training sensing outputs 603 and the first sensing labels 604. The learning apparatus calculates training errors corresponding to the first training sensing outputs 603, and optimizes the connection relationship of nodes in the first neural network 602 to minimize the training errors, and the first neural network 602 Can be learned. The control device may obtain the integrated epidermal content information from the first to third epidermal content information using the learned first neural network 602.

7 is a diagram illustrating neural networks employed to obtain a raw material composition ratio 713 according to an exemplary embodiment.

According to one embodiment, the control device is based on color information 701, gloss information 702, first skin quality information 703, second skin quality information 704, and integrated epidermal content information 705. , It is possible to obtain the raw material composition ratio 713. Here, the color information 701, the glossiness information 702, the first skin quality information 703, the second skin quality information 704, and the integrated epidermal content information 705 are respectively shown in FIGS. 1, 2 and 5 It can be generated by the technique described above with reference to.

According to one embodiment, the control device is based on color information 701, gloss information 702, first skin quality information 703, second skin quality information 704, and integrated epidermal content information 705. , A first integrated vector 710 may be generated. The first integration vector 710 may correspond to an input layer of the second neural network 711. The control device may obtain the first integrated output 712 by applying the first integrated vector 710 to the pre-learned second neural network 711. The first integrated output 712 may correspond to an output layer of the second neural network 711.

According to an embodiment, the control device may search and obtain the raw material composition ratio 713 from the database based on the first integrated output 712. The raw material composition ratio 713 may include the type of raw material and the content ratio thereof, and contents corresponding to the raw material information may be stored in a database in advance.

According to an embodiment, the control device may determine the type and content ratio of an optimized raw material suitable for individual skin by applying a deep learning technology.

According to an embodiment, the control device may obtain the predicted data 723 for each step based on the first integrated vector 710 and the raw material composition ratio 713. Here, the first integrated vector 710 and the raw material composition ratio 713 may be generated by the above-described technique.

According to an embodiment, the control device may generate the second integrated vector 720 based on the first integrated vector 710 and the raw material composition ratio 713. The second integration vector 720 may correspond to the input layer of the third neural network. The control device may obtain a second unified output 722 by applying the second unified vector 720 to a pre-learned third neural network. The second integrated output 722 may correspond to an output layer of the third neural network.

According to an embodiment, the control device may obtain the predicted data 723 for each step based on the second integrated output 722. When the cosmetic composition generated based on the raw material composition ratio 713 is virtually applied, the predicted data 723 for each step may be data obtained by predicting and analyzing changes in skin depending on the amount of application of the cosmetic composition once, the number of applications, and the application period. . Hereinafter, learning contents of the second neural network 711 and the third neural network will be described with reference to FIG. 8.

According to an embodiment, the control device may obtain the feedback information 724 based on the predicted data 723 for each step. The feedback information 724 is data for correcting the existing raw material composition ratio 713, and may be automatically generated by analyzing the step-by-step predicted data 723, and may be manually adjusted by a person in some cases.

According to an embodiment, the control device may predict the result of skin application from the condition of the skin and the composition of the cosmetic composition by applying the deep learning technology.

According to an embodiment, the control device may correct the raw material composition ratio 713 based on the feedback information 724. The control device can prepare a raw material based on the corrected raw material composition ratio 713. The control device may mix the prepared raw materials, and package the cosmetic composition thus mixed in a container prepared in advance.

According to an embodiment, the control device may store the second integrated vector 720 and the feedback information 724, which may include storing it in its own database and a blockchain network. A detailed description of the storage of the second unified vector 720 and the feedback information 724 will be described later with reference to FIG. 9.

8 is a diagram for explaining a learning method employed to obtain a raw material composition ratio according to an embodiment.

According to an embodiment, the learning device determines the raw material composition ratio from color information, gloss information, first skin quality information, second skin quality information, and integrated epidermal content information, and generates predicted data for each step to correct the neural Networks can be trained. The learning device may be a separate subject different from the control device, but is not limited thereto, and contents overlapping with the above-described learning method are omitted.

According to an embodiment, the learning device includes labeled training color information 801, labeled training gloss information 802, labeled training first skin quality information 803, and labeled training second skin quality. Information 804 and labeled training integrated epidermal containing information 805 may be obtained. The learning device may acquire training color information, training gloss information, training first skin quality information, training second skin quality information, and training integrated epidermal containing information, respectively, pre-labeled information for acquisition. Color information, training gloss information, training first skin quality information, training second skin quality information, and training integrated skin content information may be labeled according to pre-classified raw material composition ratios 813. The learning data used at this time are 10,000 pairs of color information, gloss information, first skin quality information, second skin quality information, and integrated epidermal content information, and the composition ratio of raw materials generated by skin experts. It may be determined according to the 813.

According to an embodiment, the learning device includes labeled training color information 801, labeled training gloss information 802, labeled training first skin quality information 803, and labeled training second skin quality. The first training integration vectors 810 may be generated based on the information 804 and the labeled training integrated skin-containing information 805.

According to an embodiment, the learning device may obtain the first training integration outputs 812 by applying the first training integration vectors 810 to the second neural network 811. The learning apparatus may train the second neural network 811 based on the first integrated training outputs 812 and the first integrated labels 813. The learning apparatus calculates training errors corresponding to the first training integration outputs 812 and optimizes the connection relationship between nodes in the second neural network 811 to minimize the training errors, and the second neural network 811 Can be learned. The control device may acquire information for the composition ratio of raw materials from color information, gloss information, first skin quality information, second skin quality information, and integrated epidermal content information using the second neural network 811 that has been learned. .

According to an embodiment, the learning apparatus may acquire predicted data for each step based on the first training integration vectors 810 and the raw material composition ratios 813. The learning device may obtain pre-labeled information on the first training integration vectors 810 and raw material composition ratios 813, and the labeling includes training color information, training gloss information, and training first skin quality information. , Training second skin quality information and training integrated epidermal containing information may be pre-labeled, respectively. Training color information, training gloss information, training first skin quality information, training second skin quality information, and training integrated epidermal content information are not only pre-classified raw material composition ratios 813 but also predicted results of applying the same Can be labeled together. The learning data used at this time are 10,000 pairs of color information, gloss information, first skin quality information, second skin quality information, and integrated epidermal content information, and the composition ratio of raw materials generated by skin experts. It may be determined according to the s 813 and expected effects when applied to the corresponding skin.

According to an embodiment, the learning apparatus may generate the second training integration vectors 820 based on the first training integration vectors 810 and the raw material composition ratios 813.

According to an embodiment, the learning apparatus may obtain the second training integration outputs 822 by applying the second training integration vectors 820 to the third neural network 821. The learning apparatus may train the third neural network 821 based on the second training unified outputs 822 and the second unified labels 823. The learning apparatus calculates training errors corresponding to the second training integration outputs 822 and optimizes the connection relationship between nodes in the third neural network 821 to minimize the training errors, and the third neural network 821 Can be learned. The control device may obtain step-by-step predicted data from the first integrated vector and the raw material composition ratio by using the learned third neural network 821.

9 is a diagram illustrating a block chain network according to an embodiment.

Referring to FIG. 9, in a method for the control device to store signals, the control device may store a second integrated vector and feedback information in a blockchain network. The blockchain network includes blocks 901 including a second integrated vector and feedback information; Chains 902 connecting each block 901 in chronological order; And it may be a private blockchain network including first to third network storage devices 910, 920, and 930 that store each block chain.

According to an embodiment, the block 901 of the blockchain network may include a database including the second integrated vector and feedback information. Each block 901 is generally connected in chronological order, and accordingly, new blocks 901 may be produced at 10 minute intervals. The contents of the already generated block 901 are stored in all network storage devices, and have a structure that cannot be changed unless a majority of the contents are changed within time. For example, if a block chain network having a total of 3 network storage devices has one block 901 for each network storage device, if the contents of two or more blocks 901 cannot be changed within a limited time, Each block 901 may change the value of the block 901 having contents different from the majority through verification to be equal to the majority. Accordingly, high security can be maintained, and since the number of network storage devices actually participating in the blockchain network can reach tens to hundreds of thousands of devices, higher security can be exhibited.

Chains 902 according to an embodiment may be configured with a hash value. The chains 902 allow the blocks 901 to be consecutive in time order, and in this case, each block 901 may be connected using a hash value. The block 901 may store a second unified vector, feedback information, a hash value of the second unified vector, a hash value of the feedback information, a previous header, and a header of the current block. Here, since the header of the current block functions as the previous header in the next block, the blocks 901 can be organically connected. In addition, if the content of the block changes even a little, the hash value is transformed into a completely different form, so that attempts to change the content of the block 901 can be effectively prevented. Since the header of the current block becomes the hash value of the second unified vector, the feedback information, the hash value of the second unified vector, the hash value of the feedback information, and the hash value of the total sum including the previous header, the content and the hash value of the block 901 are effectively Any attempt to compromise security by modifying it becomes difficult to succeed. This is because the moment the content and the hash value of the block 901 are modified, the content of the header also changes, and accordingly, the previous header that enters the next block also changes, and accordingly, the header of the block also changes and the previous header of the next block is changed again. I can. That is, all subsequent blocks 901 must be hacked. Therefore, it is possible to increase the security of the blockchain by using the chain 902 through the hash value.

According to an embodiment, the network storage devices include a first network storage device 910 including a raw material manufacturer of a cosmetic composition; A second network storage device 920 including a manufacturer for manufacturing a cosmetic composition; A third network storage device 930 including a distributor of the cosmetic composition; And an Internet connection network 903 connecting each of the network storage devices. The number of network storage devices classified as first, second, and third may be determined according to the number of practitioners, the number of users, and the number of storage devices included.

According to an embodiment, the first network storage device 910 may include a raw material manufacturer of a cosmetic composition, and the raw material manufacturers of the skin testing method and the cosmetic composition manufacturing method using the same may be one or more. The number of first network storage devices 910 may be determined according to the number of storage devices used by each company. The raw material manufacturer of the first network storage device 910 may use the information acquired through the blockchain network for raw material production, and may provide information on the produced raw material through the blockchain network.

According to an embodiment, the second network storage device 920 may include a manufacturer that manufactures a cosmetic composition. A manufacturer of the second network storage device 920 may include a user of a method for manufacturing a cosmetic composition, and the number of the second network storage device 920 may be determined according to the number. The manufacturer of the second network storage device 920 may facilitate learning of the convolutional neural network and the neural network by manually inputting information on errors occurring during the execution of the cosmetic composition manufacturing method.

According to an embodiment, the third network storage device 930 may include a distributor of the cosmetic composition. The third network storage device 930 can browse and store the blockchain network according to the permission of the user of the second network storage device 920, and in some cases, the third network storage device 930 of the blockchain network You may lose your role. The distributor of the third network storage device 930 can perform a task of distributing the cosmetic composition that has been manufactured to a target vendor, and can confirm the completion of manufacturing the cosmetic composition through a blockchain network.

The Internet connection network 903 according to an embodiment generally refers to an Internet connection network exhibiting a speed of 10 Mb/s or more, and is a connection network including wired, wireless, optical cable technology, etc., and is a local area network (LAN) or wireless local area network (LAN). Area Network), Wide Area Network (WAN), Personal Area Network (PAN), and the like, but are not limited thereto.

According to an embodiment, a method for manufacturing a cosmetic composition is intended to convert information into big data through a blockchain network and use enhanced security.

10 is an exemplary diagram of a configuration of an apparatus according to an embodiment.

The device 1201 according to an embodiment includes a processor 1202 and a memory 1203. The device 1201 according to an embodiment may be the above-described server or terminal. The processor may include at least one of the devices described above with reference to FIGS. 1 to 9, or may perform at least one of the methods described above with reference to FIGS. 1 to 9. The memory 1203 may store information related to the above-described method or a program in which the above-described method is implemented. The memory 1203 may be a volatile memory or a nonvolatile memory.

According to an embodiment, the processor 1202 may execute a program and control the device 1201. The code of a program executed by the processor 1202 may be stored in the memory 1203. The device 1201 is connected to an external device (for example, a personal computer or a network) through an input/output device (not shown), and may exchange data.

The embodiments described above may be implemented as a hardware component, a software component, and/or a combination of a hardware component and a software component. For example, the devices, methods, and components described in the embodiments include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions, such as one or more general purpose computers or special purpose computers. The processing device may execute an operating system (OS) and one or more software applications executed on the operating system. In addition, the processing device may access, store, manipulate, process, and generate data in response to the execution of software. For the convenience of understanding, although it is sometimes described that one processing device is used, one of ordinary skill in the art, the processing device is a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as a parallel processor.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of the program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

The software may include a computer program, code, instructions, or a combination of one or more of these, configuring the processing unit to behave as desired or processed independently or collectively. You can command the device. Software and/or data may be interpreted by a processing device or to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodyed in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

As described above, although the embodiments have been described by the limited drawings, a person of ordinary skill in the art can apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order from the described method, and/or components such as a system, structure, device, circuit, etc. described are combined or combined in a form different from the described method, or other components Alternatively, even if substituted or substituted by an equivalent, an appropriate result can be achieved.

Therefore, other implementations, other embodiments and claims and equivalents fall within the scope of the following claims.

Claims (5)

In the method of manufacturing a cosmetic composition based on artificial intelligence,
Controlling a camera in the skin analyzer under a first light source and a second light source of a skin analyzer to obtain a first face image and a second face image, respectively;
Generating a first image vector based on the first face image and the second face image;
Applying the first image vector to a pre-trained first convolutional neural network to obtain a first image output;
Based on the output of the first image, the skin range-the skin range includes the first and second face images, and the forehead and cheeks recognized as skin in the first and second face images. Obtaining-means range;
Obtaining color information and gloss information within the skin range, based on the skin range;
Generating a second image vector based on the skin range;
Applying the second image vector to a pre-trained second convolutional neural network and a third convolutional neural network to obtain a second image output and a third image output, respectively;
Based on the output of the second image, the first skin quality information-the first skin quality information includes structures to be improved within a range of skin that can be primarily distinguished from the face, and the structures to be improved include acne, wrinkles, gombo, and atopy. Dermatitis, erythema, urticaria and eczematous skin disease means-obtaining; And
Acquiring second skin quality information based on the output of the third image-the second skin quality information means information on particles of the epidermis capable of expressing the overall management status of the face
Including,
The first light source is
It is yellow light in the wavelength range of 576 to 580 nm,
The second light source is
Cyan light in the wavelength range of 490 to 520 nm,
The above color information is
It means graphic information expressed in 8-bit 256-color RGB color for each pixel of the skin range,
The gloss information is
Means a numerical value expressing the intensity of light when the reflected light is received at the same angle as the incident light on the surface of the skin range,
The first convolutional neural network is
Acquiring labeled training first face images and labeled training second face images;
Generating first training image vectors based on the labeled training first face images and the labeled training second face images;
Applying the first training image vectors to the first convolutional neural network to obtain first training image outputs; And
Training the first convolutional neural network based on the first training image outputs and first image labels
Including more,
The second convolutional neural network and the third convolutional neural network are
Obtaining labeled training skin ranges;
Based on the labeled training skin ranges, generating second training image vectors;
Applying the second training image vectors to the second convolutional neural network and the third convolutional neural network to obtain second training image outputs and third training image outputs; And
Training the second convolutional neural network and the third convolutional neural network based on the second training image outputs, the third training image outputs, second image labels, and third image labels
Including more,
The first image labels include skin ranges corresponding to 10,000 pairs of labeled training first face images and labeled training second face images generated by skin experts,
The second image labels are
Including first skin quality information corresponding to 10,000 skin ranges generated by the skin experts,
The third image labels are
Including second skin quality information corresponding to 10,000 skin ranges generated by the skin experts
Cosmetic composition manufacturing method.
The method of claim 1,
The first to third epidermal content information by applying the cotton for examination rubbed for 30 seconds to the face region corresponding to the first skin region, the second skin region, and the third skin region, each of which has been subdivided into the skin range, is applied to the test kit. Obtaining the epidermal content information including the content and content ratio of oil, moisture and contaminants that can be obtained from the epidermis;
Generating a first sensing vector based on the first to third epidermal content information;
Obtaining a first sensing output by applying the first sensing vector to a first neural network; And
Based on the first sensing output, acquiring integrated skin content information
Including more,
The first neural network is
Acquiring labeled training first to third epidermal containing information based on facial regions corresponding to the first to third skin regions of the skin ranges;
Generating first training sensing vectors based on the labeled training first to third epidermal containing information;
Applying the first training sensing vectors to the first neural network to obtain first training sensing outputs; And
Training the first neural network based on the first training sensing outputs and first sensing labels
Including more,
The first sensing labels are
Including integrated epidermal containing information corresponding to 10,000 pairs of the first to third epidermal containing information generated by the skin experts
Cosmetic composition manufacturing method.
The method of claim 2,
Preparing a raw material based on data including the color information, the gloss information, the first skin quality information, the second skin quality information, and the integrated epidermal content information;
Mixing the prepared raw materials to prepare a cosmetic composition; And
Packaging the prepared cosmetic composition
Further comprising
Cosmetic composition manufacturing method.
The method of claim 3,
The step of preparing the raw material
Generating a first integrated vector based on the color information, the gloss information, the first skin quality information, the second skin quality information, and the integrated epidermal content information;
Applying the first integrated vector to a second neural network to obtain a first integrated output;
Obtaining a raw material composition ratio based on the first integrated output;
Generating a second integrated vector based on the first integrated vector and the raw material composition ratio;
Applying the second unified vector to a third neural network to obtain a second unified output;
Based on the second integrated output, step-by-step predicted data-the step-by-step predicted data predicts the change in skin according to the amount of one application of the cosmetic composition, the number of applications, and the application period when applying the cosmetic composition generated based on the raw material composition ratio. Obtaining the analyzed data;
Obtaining feedback information based on the step-by-step predicted data;
Correcting the raw material composition ratio based on the feedback information;
Preparing the raw material based on the modified raw material composition ratio; And
Storing the second unified vector and the feedback information
Including,
The second neural network is
Acquiring labeled training color information, labeled training gloss information, labeled training first skin quality information, labeled training second skin quality information, and labeled training integrated epidermal content information;
Based on the labeled training color information, the labeled training gloss information, the labeled training first skin quality information, the labeled training second skin quality information, and the labeled training integrated epidermal content information Generating first training integration vectors;
Applying the first training integration vectors to the second neural network to obtain first training integration outputs; And
Training the second neural network based on the first training unified outputs and first unified labels
Including more,
The third neural network is
Obtaining raw material composition ratios based on the first training integration outputs;
Generating second training integration vectors based on the first training integration vectors and the raw material composition ratios;
Applying the second training integration vectors to the third neural network to obtain second training integration outputs; And
Training the third neural network based on the second training unified outputs and second unified labels
Including more,
The first unified labels are
Including raw material composition ratios corresponding to 10,000 pairs of color information, gloss information, first skin quality information, second skin quality information, and integrated epidermal content information generated by skin experts,
The second unified labels are
Raw material composition ratios corresponding to 10,000 pairs of color information, gloss information, first skin quality information, second skin quality information and integrated epidermal content information generated by skin experts, and applying the same to the skin Including the expected effects of
Cosmetic composition manufacturing method.
The method of claim 4,
Storing the second integration vector and the feedback information
Storing the second integrated vector and the feedback information in a blockchain network,
The blockchain network is
Blocks including the second integration vector and the feedback information;
Chains connecting the blocks in chronological order; And
Network storage devices that store the blocks and the chains
It is a private blockchain network that includes,
The network storage devices
A first network storage device including a manufacturer of raw materials for the cosmetic composition;
A second network storage device including a manufacturer that manufactures the cosmetic composition;
A third network storage device including a distributor of the cosmetic composition; And
Internet connection network connecting each network storage device
Containing
Cosmetic composition manufacturing method.

Images