KR102288001B1 - Device for generating job image having face to which age transformation is applied and photo booth including the same - Google Patents

Abstract

Embodiments of the present invention relate to a job image generating device with an age-converted face and a photo booth including the same, which have a process including the steps of: acquiring an original image including a subject's face at an original age; obtaining original age information of a subject; determining a background image in which a job is expressed for use for generating a job image; converting the face of the subject at the original age into the face of the subject at a target age; and generating the job image of the target by synthesizing the age-converted face with a determined face region of the background image. Accordingly, an age-converted image having an adult-age face may be generated based on a generated texture and shape.

Description

Occupational image generating device with age-converted face and photo booth including the same

Embodiments of the present invention relate to image processing technology, and more particularly, to generate an age-converted image having a face expected to have when the subject becomes a target age from a face image of a subject at a specific age, The present invention relates to a job image generating apparatus and a system (eg, a photo booth) including the same for indicating a case in which the target has a job at a target age by using the target's face at the age.

Occupation is a social activity with various and important meanings ranging from economic meaning in an individual's life to the meaning of self-realization through professional life.

In recent years, advancement and segmentation have been progressing in all fields of occupation, and the earlier the preparation period for career selection, the more advantageous it is. Career selection should be made carefully, based on a wealth of knowledge and information on technological/social trends. In general, children of infants and adolescence have relatively low ability to analyze these technological/social trends, so parental advice has a great influence on career choices.

However, the subject of career choice is the child himself/herself, and therefore, it is highly probable that the child's active determination of the occupation will enable the child to make an effort to acquire the corresponding occupation experienced in the process of growth.

In the end, it will be the best vocational education for parents to enable diverse and specific job information to enable active career choices for young children.

Imagining the future of your child when he or she has the job is the most sure motivator for choosing a job. However, in general, the job information provided by the parent only provides information about the job itself or information about others who have the job, so there is a limit to the child's ability to accept and empathize with the parent's job information.

Patent Publication No. 10-1998-0065049 Patent Laid-Open Publication No. 10-2002-0007744

Goodfellow, Ian J.; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley, David; Ozair, Sherjil; Courville, Aaron; Bengio, Yoshua (2014). "Generative Adversarial Networks" Mehdi Mirza; Simon Osindero, (2014) “Conditional Generative Adversarial Nets”

According to an aspect of the present invention, an age conversion image is generated by converting the face of a target included in the original image into a target age, and the face at the target age included in the age conversion image is synthesized into a background image expressing a job To provide a device for generating a job image having an age-converted face, indicating a case in which the subject has reached the target age and has a corresponding job.

In addition to this, it is possible to provide a system to which the device is applied, such as a photo booth and a kiosk.

An apparatus for generating a job image having an age-converted face, including a processor according to an aspect of the present invention, includes: obtaining an original image including a face of a subject at an original age; obtaining original age information of the subject; determining a background image in which a job is expressed for use in generating a job image; converting the face of the subject at the original age into the face of the subject at the target age; and synthesizing the age-converted face with the determined face region of the background image to generate a job image of the target.

In one embodiment, the device may be further configured to perform the step of: providing a career guidance service image including at least a vocational image among the original image, the background image, the age conversion image and the vocational image to the user.

In an embodiment, the determining of the background image for the job image may include: providing an interface screen for inducing a user's input for selecting a background image by displaying at least a portion of a previously stored candidate background image; and when receiving an input for selecting a background image, determining a candidate background image corresponding to the input as a background image for a job image of the target. Here, the candidate background image is configured to express the characteristics of the job, and includes at least a part of a face of a person having a corresponding job, which is different from the target.

In one embodiment, the step of determining the background image for the job image, before providing the interface screen inducing the user's input to select the background image, to display a job-related list to select a desired job It may further include the step of providing a selection screen. Here, the occupation-related list includes at least one of one or more occupational items (listing) and one or more occupational group items.

In an embodiment, the interface screen includes at least a portion of a candidate background image related to the gender of the subject when the subject's gender information is obtained.

In an embodiment, the converting into the target's face at the target age includes: extracting a landmark from the target's face of the original image; generating a facial texture of the target at the target age from the target's face of the original image from which the landmark is extracted; generating a face shape of the subject at the target age from the face of the subject of the original image from which the landmark is extracted; and generating an age-converted face of the target based on a facial texture and a face shape of the target at the target age.

In an embodiment, the generating of the facial texture at the target age includes: generating a shapeless facial texture from the face of the target of the original image from which the landmark is extracted; and applying the shapeless facial texture to a pre-learned texture transformation model to generate a shapeless facial texture of the target at the target age.

In an embodiment, the texture transformation model may include: a first generator configured to output transformed data of a second domain by applying noise to input data of a first domain; and a second generator configured to output converted data of a third domain by applying noise to input data of the first domain. Here, each generator is configured to be converted into input data of the first domain upon re-conversion of the converted data into data into the first domain.

In an embodiment, the texture transformation model may include a generator that applies noise and condition information to input data of a first domain to output transformation data of a plurality of different domains including a second domain and a third domain. .

In an embodiment, the texture transformation model is a pre-trained model using a plurality of training samples to output data corresponding to a facial texture at the target age, and each training sample is a model of a training target at the target age. It may include label data including a face texture and the gender of the training target.

In an embodiment, the facial texture of the training target may be a shapeless facial texture obtained from the training target's face.

In an embodiment, the generating of the face shape of the subject at the target age includes: extracting facial shape features of the subject of the original image based on the landmark of the face of the original image; generating facial features of the target at the target age by applying the facial shape features of the original image to a pre-learned shape transformation model; and reconstructing the facial shape of the target at the target age based on the facial shape features of the target at the target age.

In an embodiment, the shape transformation model is generated by modeling a relationship between age and a facial shape feature at the corresponding age, the difference between the age function value at the target age and the age function value at the original age, and the It may be modeled to calculate the facial shape characteristics of the target at the target age based on the facial shape characteristics of the target of the original image.

In one embodiment, the shape transformation model is a model pre-trained using a plurality of training samples and label data indicating the target age to output the facial shape features at the target age, wherein the training samples in each set are corresponding It may include the facial shape features of the training target at age.

In an embodiment, when the shape transformation model is configured to output facial shape features at any one of a plurality of ages, the shape transformation model is a pre-trained model using a plurality of sets of training samples, each set may include a plurality of training samples at a specific age among the plurality of ages and label data indicating the specific age, and each training sample in each set may include a facial shape feature of a training target at the corresponding age.

In an embodiment, the shape transformation model may be modeled based on an age function for each facial shape feature when the dimension of the facial shape feature is N (here, N is an integer greater than or equal to 1).

In an embodiment, the generating of the age-converted face of the target includes warping the face texture at the target age with the face shape of the target at the target age, thereby converting the age of the face at the target age. It may include generating as a face.

In one embodiment, the generating of the job image of the target may include: extracting each of the landmarks of the face of the background image and the age-converted face; mapping the landmark of the age-converted face to the face region of the background image based on each extracted landmark; warping the face texture of the age conversion image based on the location of the landmark mapped to the background image; warping the face texture of the background image based on the location of the landmark mapped to the background image; generating a composite area mask for filtering an inner area based on the moved landmark of the face of the background image; and filtering the face region of the target of the warped age transformation image using the composite region mask, and transplanting the filtered face region to the face region of the background image.

In an embodiment, the mapping may be performed based on anatomical facial features that each landmark means.

In an embodiment, the generating of the job image may include synthesizing the age-converted face with at least one of the plurality of frames when the selected background image is a moving picture composed of a plurality of frames.

In an embodiment, the generating of the job image may further include generating a job video composed of frames synthesized to have an age-converted face.

In an embodiment, the device is one of displaying the career guidance service image, printing the career guidance service image, and transmitting the career guidance service image to provide the career guidance service image It is configured to do more than

A photobooth according to another aspect of the present invention includes an apparatus according to the above-described embodiments; and a booth at least partially surrounded by a wall to form a space in which the object can be located.

An apparatus for generating a job image according to an aspect of the present invention generates a texture at an adult age from a face image of a child using a pre-learned texture transformation model, and uses a pre-learned age function from the face image of a child After generating the shape at the adult age, an age conversion image having the face at the adult age may be generated based on the generated texture and shape.

In addition, in synthesizing the age conversion image with the occupational image, a more realistic and user-customized image may be provided by naturally synthesizing the face region of the age transformation image with the face region of the occupational image.

As a result, it is possible to provide concrete and realistic images that support imagining a child's desired career path, and also support career decision-making.

Effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

In order to more clearly explain the technical solutions of the embodiments of the present invention or the prior art, drawings necessary for the description of the embodiments are briefly introduced below. It should be understood that the drawings below are for the purpose of explaining the embodiments of the present specification and not for the purpose of limitation. In addition, some elements to which various modifications such as exaggeration and omission have been applied may be shown in the drawings below for clarity of description.
1A and 1B are diagrams conceptually illustrating an operation of an apparatus for generating an occupational image according to an embodiment of the present invention.
2 is a conceptual configuration diagram of an apparatus for generating an occupational image according to an embodiment of the present invention.
3 is a diagram illustrating an interface screen for selecting a background image, according to an embodiment.
4 is a view for explaining the operation of the age conversion unit, according to an embodiment of the present invention.
5 is a diagram for explaining a process of determining an age function for a facial shape feature, according to an embodiment of the present invention.
FIG. 6 is a diagram illustrating an age function for a second shape feature trained by the process of FIG. 5 .
7 is a conceptual diagram of a photobooth system according to an embodiment of the present invention.
8 is a flowchart of a method for generating a job image, according to an embodiment of the present invention.
9 is a diagram exemplarily showing a result according to an age conversion operation, according to an embodiment of the present invention.
10 is a flowchart of an image synthesis process according to an embodiment of the present invention.
11 is a diagram illustrating a landmark extraction result of an age conversion image and a background image according to an embodiment of the present invention.
12 is a diagram illustrating a landmark of an age conversion image mapped to a background image, according to an embodiment of the present invention.
13 is a diagram illustrating a result of moving a landmark of a face of a background image according to an embodiment of the present invention.
14 is a diagram illustrating a result of positional movement of a landmark of a face of an age conversion image according to an embodiment of the present invention.
15 is a diagram illustrating a composite area mask according to an embodiment of the present invention.
16 is a diagram illustrating a composite image when a target job is a police officer according to an embodiment of the present invention.
17 is a diagram illustrating a composite image when a target job is a doctor, according to an embodiment of the present invention.
18A and 18B are diagrams illustrating a career guidance service image according to embodiments of the present invention.

The terminology used herein is for the purpose of referring to specific embodiments only, and is not intended to limit the invention. As used herein, the singular forms also include the plural forms unless the phrases clearly indicate the opposite. The meaning of "comprising," as used herein, is intended to specify a particular characteristic, region, integer, step, operation, element, component, and/or part, and is intended to specify another characteristic, region, integer, step, operation, element, component and/or component. or the presence or addition of parts.

Although not defined otherwise, all terms including technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present invention belongs. Commonly used terms defined in the dictionary are additionally interpreted as having a meaning consistent with the related technical literature and the presently disclosed content, and unless defined, they are not interpreted in an ideal or very formal meaning.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

1A and 1B are diagrams conceptually illustrating an operation of an apparatus for generating an occupational image according to an embodiment of the present invention.

According to an embodiment of the present invention, the occupational image generating apparatus is configured to age-convert the face of the target included in the original image. The age conversion process refers to an image processing operation of converting the face of the subject at the age of the subject (eg, the age of the subject at the time of image shooting) into a face expected to be possessed by the subject when the user reaches an age desired to be converted. do. The occupational image generating apparatus generates an age-converted image having a face of a target at a target age by performing an age conversion process. Then, the job image generating device synthesizes the face of the target at the target age included in the age conversion image with the background image in which the job is expressed to generate a job image expressing the case where the target has a corresponding job at the target age do.

Here, the target is a target whose face is age-converted, and in one embodiment, the target refers to a child as shown in FIG. 1 . As used herein, a child refers to a person who has not reached the age of majority. Children include infants, toddlers, and adolescents.

In this specification, a user is a person who uses the device 1 and a system including the same, and is an object or a third party related to the object. The third party may be, for example, a guardian protecting the object.

The original image is an image of the subject's face that has not yet been processed for age conversion. The face of the original image is the original face of the object at the time of image acquisition. The age of the object corresponding to the face of the original image is the age before the conversion process and is referred to as the original age. For example, the age of the subject included in the original image is the age at which the subject was photographed.

In this specification, the target age indicates an age at which a user's desired age-converted face is generated. The target age may be an age desired by a subject (eg, a child) or a user (eg, a parent) for the job image, or an age preset in the device 1 .

In an embodiment, the target age may be an age different from the age at which the subject was photographed, such as an age corresponding to when a child grows up and becomes an adult as shown in FIG. 1 .

As used herein, an adult generally refers to a person who has reached the age of majority. The age corresponding to an adult is not limited to a specific age after the age of majority, and may be referred to as a range including one or more ages. Also, in certain embodiments, the age corresponding to an adult may be different for each occupation.

2 is a conceptual configuration diagram of an apparatus for generating an occupational image according to an embodiment of the present invention.

Referring to FIG. 2 , the apparatus for generating a job image 1 includes a control unit 50; Age conversion unit 60; and an image synthesizing unit 70 . In some embodiments, the device 1 includes an imaging device 9; input device 10; Transceiver 20; memory 30; and a display device 40; And it may further include at least a portion of the image printing device (80).

The device 1 according to embodiments may have aspects that are entirely hardware, entirely software, or partly hardware and partly software. For example, the system may collectively refer to hardware equipped with data processing capability and operating software for driving the same. As used herein, terms such as “unit”, “module,” “device,” or “system” are intended to refer to a combination of hardware and software run by the hardware. For example, the hardware may be a computing device capable of processing data including a central processing unit (CPU), a graphic processing unit (GPU), or another processor. In addition, software may refer to a running process, an object, an executable file, a thread of execution, a program, and the like.

The photographing device 9 is configured to photograph a subject to obtain an image of the subject. The photographing device 9 includes, for example, various photographing devices, such as an image sensor, a camera, or a video camera. The camera may include an instant camera, such as a Polaroid camera.

The input device 10 is configured to receive data, information commands, etc. related to the operation of generating the occupational image. The input device 10 includes, for example, a mouse, a keyboard, a microphone, a touch sensor, a gesture sensor, etc., but is not limited thereto, and may be implemented as various devices capable of receiving a user's input.

The transceiver 20 is configured to transmit/receive data through a wired/wireless electrical connection with a user's mobile device. Here, the mobile device is an electronic device that a user can carry, and includes, for example, a smart phone, a cellular phone, smart glasses, a smart watch, a wearable device, a digital camera, a tablet, and the like. The wired/wireless electrical connection includes, for example, various methods, such as 3G, 4G, 5G, LTE, WiFi, Bluetooth, and the like.

The memory 30 may store data used for age conversion or job image generation. The data used for age transformation includes a basic image of age transformation, a model for age transformation, and the like, such as the previously obtained original image. The data used for generating the job image includes a background image for each job, a model for generating the job image, and the like.

The display device 40 outputs information stored and/or processed by the providing device 1 . In one embodiment, the display device 40 may display an image stored in the device 1 or an image processed by the device 1 . In addition, a user interface for receiving a user input for generating a job image may be further displayed.

The display device 40 may include various display devices such as an LCD, an OLED, a flexible screen, and the like.

Although the input unit 10 and the display device 40 are separated in FIG. 2 , in an embodiment, the display device 40 and the input unit 10 may be implemented as one component to receive input and output information. there is. For example, the input unit 10 and the display device 40 may be implemented as a touch screen in which the touch panel forms a layer structure with the screen.

The control unit 50 controls the overall operation of the device 1 . For example, the control unit 50 converts the information obtained by the interaction between the input device 10 and the user and/or the information transmitted/received from the external device (eg, the user's mobile device) to the age conversion unit. Provided to ( 60 ), the device ( 1 ) generates an age-converted image, or provides it to the image synthesis unit ( 70 ) to generate an occupational image.

When the age conversion unit 60 receives the original image, the original age information, and the target age information, the age conversion unit 60 converts the age of the original image to the face of the target included in the original image. and to generate an age-converted image having a face expected to have when the subject becomes a target age.

In addition, the image synthesizing unit 70 is configured to generate a job image expressing the job while having a face at the target age based on the age conversion image and the background image (eg, generated by the age converting unit 70). .

The operation of the age converting unit 60 and the image synthesizing unit 70 will be described in more detail with reference to FIG. 4 below.

The image printing device 80 is configured to output the image obtained by the device 1 through printing, including an original image, an age conversion image, a background image, a job image, and the like.

The printing includes printing through paper, printing through photos (ie, prints), and printing through plastics. The image printing device 80 may include, for example, a black-and-white printer that receives image data and prints it on paper, a color printer, a 3D printer that prints on plastic, and/or (eg, a Polaroid instant camera, etc.) Including a photo printing device, the image printing device 80 is not limited to the above-described devices.

The device 1 acquires an original image to be subjected to age conversion processing by an image acquisition device (eg, the photographing device 9 or the transceiver 20). For example, the device 1 acquires an original image by photographing the face of the target by the photographing device 9 or an external device (eg, the user's mobile device) by the transceiver 20 . or an external computer) through electrical communication to receive the original image.

The device 1 is configured to determine the original age of the subject. Also, the device 1 may further determine the gender of the subject.

In one embodiment, the device 1 determines the original age and/or gender of the subject based on a user input including original age information and/or gender information (eg, obtained by the input device 10 ). . The user input may be an input for details of an object. In some embodiments, the user input for details of the subject further includes the subject's name, identification number, and the like.

In another embodiment, the device 1 receives only the original image, analyzes the subject's face of the original image, calculates the age value of the subject, and generates a job image based on the calculated age value of the subject You may. In addition, the gender of the subject may be further determined by analyzing the subject's face in the original image. The calculation of the age value and the determination of the gender of the subject may be performed based on at least a part of the analysis for age transformation.

In another embodiment, the device 1 may determine the original age and gender of the subject based on the above-described user input and original image analysis. For example, the original age is obtained by a user input, and the gender is obtained based on the analysis of the original image.

For example, the device 1 is configured to apply an input image (ie, an original image) to a CNN-based deep learning network to calculate age and/or gender information in a face region. The CNN-based deep learning network includes a plurality of convolutional layers; As a structure including a fully connected layer, age and/or gender are recognized using the extracted age feature and a classifier or regressor.

The CNN-based deep learning network is trained on a set of training samples including face images and one or more label data. Each training sample in the set of training samples includes label data for age, and/or label data for gender, respectively.

In some embodiments, the apparatus 1 may further perform an operation of detecting a face region in an input image (ie, an original image) and/or a normalization operation. Only the face region can be applied to the CNN-based deep learning network by this preprocessing process.

Also, the device 1 may be configured to receive an input for selecting a background image in which a job is expressed. The background image is configured to express the characteristics of the job, and may include at least a part of a face of a person having a corresponding job, which is different from the target. For example, the background image may be a picture of another person wearing a uniform (eg, a police uniform or a doctor's gown) for a corresponding job.

In an embodiment, the device 1 may determine a background image for the job image of the target based on an input for selecting the background image. To this end, the device 1 is configured to provide an interface screen for selecting a background image from among one or more candidate background images in which a job is expressed.

3 is a diagram illustrating an interface screen for selecting a background image, according to an embodiment.

The interface screen of FIG. 3 induces a user's input for determining the background image by displaying one or more candidate background images. When a user input is received, the device 1 determines a candidate background image corresponding to the input as a background image for a job image.

A part of the interface screen may be displayed to the user according to the resolution and display size of the display device 40 . Other parts that are not displayed to the user are further provided to the user by the screen moving command.

The candidate background image is a candidate image that can be selected as a background image to be used to generate a job image. The background image candidate may be at least one portrait of a person representing a specific occupation group (or occupation). For example, as shown in FIG. 3 , the background image candidates include photos of celebrities and world greats. In the device 1, the candidate background image is previously associated with a specific job. Also, the candidate background image may be further related in advance to a job group to which a specific job belongs. In addition, the candidate background image may be further related to the main gender of the related job. That is, the candidate background image is related to at least a job among a job, a job group, and a gender, and may be classified or searched based on the related element.

As shown in FIG. 3, the interface screen for the background image has various structures.

In one example, the interface screen for the background image is configured to include at least a portion of the candidate background images stored in advance.

In another example, the interface screen for the background image is configured to include at least some of the candidate background images related to the selected job (or job group) after the user's job (or job group) is selected.

Then, the device 1 is further configured to provide a selection screen for selecting a desired job by displaying a job-related list before providing the interface screen for determining the background image. The job-related listing includes one or more occupational listings and/or one or more occupational group entries. When receiving a selection input for a specific job item, the device 1 provides the interface screen displaying one or more candidate background images related to the specific job item. When the device 1 receives a selection input for a specific occupational group item, it provides a sub-list including the occupational item included in the corresponding occupational group, or one or more candidate background images related to the occupation included in the corresponding occupational group. and provides the interface screen displaying

Alternatively, the device 1 may be further configured to automatically determine a predetermined representative image for the occupation (or occupation group) selected by the user from the occupation-related list as the background image.

In another example, the interface screen for the background image is configured based on gender. Hairstyles, clothes, etc. of the background image may be different depending on gender, so selecting a background image showing the same gender as the subject's gender is required to generate a natural synthetic image.

When the subject's gender information is obtained before providing the interface screen for determining the background image, the device 1 selects a candidate background image preset to be related to the corresponding gender. The device 1 configures the interface screen with a candidate background image related to the corresponding gender and provides it to the user. That is, it is further configured to filter the candidate background image related in advance to the determined gender of the user, and to provide at least a part of the filtered candidate background image. For example, when the gender of the subject is determined to be male, the device 1 provides an interface screen including a background image of a candidate related to a job determined in advance to be related to a male.

As described above, the gender of the subject may be obtained based on a user input and/or analysis of an original image. When the device 1 has previously obtained the subject's gender information, an input for selecting the gender of the background image is not separately required.

In some embodiments, when the gender of the subject is not previously obtained prior to selecting the background image, the user input for selecting the background image in the device 1 may include an input for selecting the gender. For example, the device 1 may provide an interface for selecting a job image having the selected gender after first providing an interface for selecting the gender of the background image.

Also, the device 1 may be configured to provide an interface screen according to a combination of the above-described examples. For example, the device 1 first filters a candidate background image based on occupation (or occupation group) selection and/or gender, and then interfaces to select a background image for the occupation image from among the filtered candidate background images. It may be configured to provide a screen.

In another embodiment, the device 1 may automatically determine the background image based on the original image and/or the age-converted image below.

The background image may be automatically selected based on the original image and/or the age-converted image. For example, the background image may be selected based on at least one of image characteristics of the original image and facial characteristics in the age-converted image.

The image characteristics of the original image include lighting, resolution, sharpness, and the like. The device 1 calculates the image characteristics of the original image, and selects a background image having similar characteristics as a background image for the job image. For example, the device 1 may calculate the characteristics of the original image through an image brightness analysis technique, an image quality assessment technique, or the like.

The facial characteristics of the age conversion image include, but are not limited to, a face shape (eg, a round shape, an oval shape, etc.), a hair characteristic (straight hair, curly hair, a parted direction), an ear shape, a forehead shape, a face pose, and the like.

In one embodiment, the device 1 may be configured to calculate the facial characteristics of an age conversion image to be described below, and to select a background image having similar characteristics as a background image for a job image. The apparatus 1 may calculate a face shape and a face pose based on, for example, information on feature points of a face used in age conversion. Also, the device 1 may calculate an ear shape, a forehead shape, a hair characteristic, etc. through, for example, an image segmentation technique.

When the occupational image generating device 1 receives the original image, the original age information, and the target age information, the age conversion process is performed on the original image so that the face of the target included in the original image becomes the target age. In this case, it is possible to generate an age-converted image having a face expected to have.

The age conversion unit 60 is configured to perform age conversion processing on the input image based on an age (eg, an original age) associated with the input image and a target age.

As described above, the original age in the device 1 is determined based on the acquired age information of the subject (eg, through the input device 10 or the transceiver 20), or automatically measured from the original image. can be

In an embodiment, the target age in the device 1 may be determined based on the acquired target age information (eg, through the input device 10 or the transceiver 20 ).

In another embodiment, the target age in the device 1 may be preset for each background image. For example, the age to be converted may be preset for each occupation, such as 40 years old for a doctor and 25 years old for a police officer. This is because the age of a person engaged in the relevant occupation differs according to occupation-related characteristics such as occupation group, job, occupation, years of service, annual salary, and welfare. In this case, when an input for selecting the background image is received, a preset conversion age for the background image is determined as the target age. In this case, since input of target age information is unnecessary, a separate input of target age may be omitted.

The target age may be set as an average age of each job. The device 1 may perform face transformation into a target age by storing job-related characteristics in an internal storage device or by receiving the target age from the outside.

Additionally, the target age may be plural for each job. It is a factor that greatly influences the choice of a job, and whether the job can be engaged in the long term. To this end, for each job, for example, a first target age, a second target age, and a third target age may be set. Here, the first target age represents the age of people with low rank and experience (eg, the average age of new employees or the average age of 3 years or less) in the relevant occupational worker, and the third target age is those with very high rank and experience. represents the age (eg, an average age of an executive level, an average age of 20 years or more of experience), and the second target age represents an age corresponding to workers belonging to between the first target age and the third target age.

Also, when there are a plurality of target ages for each job, the device 1 may use a background image associated with each target age to generate a job image. This is because, even within the same job, the job, clothes, environment, etc., change depending on the career. For example, when a police officer is selected as a job, the background related to the first target age is configured to represent an attire representing a rank in the first target age band, a place where a task is mainly performed when the rank is held, and the like. Occupational characteristic data for each job, including the location, position, and the like, performed mainly, uses statistically analyzed results based on various job information.

The original age or the target age may be expressed in various ways. In one example, both ages may be expressed as an age based on birth (eg, 7 years old or 25 years old). In another example, one of the two ages may be expressed as a reference age (eg, 7 years old and 7 to 18 years older).

As described above, the target age may be a value specifically expressed in units of one year, but the target age of the present invention is not limited thereto. In another embodiment, the target age may be set as an age corresponding to a group (eg, age group such as 20's, 30's, 40's, etc.) consisting of a plurality of different age groups. The age corresponding to the cluster may be a maximum value, a minimum value, a median value, and a mode value within the cluster, a representative value defined by a user, and the like.

In one embodiment, the device 1 receives a job selection command before performing face transformation by the age conversion unit 60 or (eg, via the input device 10 or the transceiver 20 ). Thereafter, a single target age or a plurality of target ages may be determined for the corresponding job.

In an embodiment, the age conversion unit 60 performs an age conversion operation based on a facial texture and shape.

4 is a view for explaining the operation of the age conversion unit, according to an embodiment of the present invention.

Referring to FIG. 4 , the age conversion unit 60: extracting a plurality of landmarks from the original face (S410), using a pre-learned texture conversion model, from the texture at the original age to the texture at the target age generating (S420); generating a shape at the target age using the pre-learned age function (S440); and generating a face at the target age based on the texture and shape at the target age ( S440 ).

The age conversion unit 60 extracts a landmark from the original face included in the original image to generate an age-converted face image (S410).

In step S410, the extraction of the landmark of the face may be performed after determining the face region corresponding to the original face in the original image. In this case, various region detection techniques for detecting a specific region from an image, such as Haar, Neural Network (NN), Support Vector Machine (SVM), Gabor, and SIFT, may be used, but the present invention is not limited thereto.

The landmark is information related to facial anatomical features such as eyes, nose, mouth, and ears. The landmark has a constant relative position within the face, and the geometric relationship according to the face pose is less changed. Here, the pose indicates a facial expression, a rotation direction of the face, an inclination angle, or the like.

In one embodiment, the landmark is extracted as a point indicating characteristics for distinguishing the face, such as the center of the eyes, the center of the nose, the end points of both mouths, the point at the position with the widest interval among the face contours, the center of the chin contour, etc. can be For example, as shown in FIG. 4 , a landmark set including 68 landmark points may be extracted.

In step S410, the landmark may be extracted by various landmark extraction algorithms capable of detecting the landmark of the face in the face region. The landmark extraction algorithm includes, for example, Active Contour Model (ACM), Active Shape Model (ASM), Active Appearance model (AAM), Supervised Descent Method (SDM), or a neural network, but is not limited thereto. . The age converter 60 may detect facial landmarks in the face region, but is not limited thereto.

The occupational image generating apparatus 1 may further acquire location information and identification information of the landmark of the original face through the age conversion unit 60 . Here, the identification information includes the definition of anatomical facial features defined by each landmark. For example, the identification information of the first landmark may include information indicating the center of the eye, and identification information of the second landmark may include information indicating the left end point of the mouth.

Secondary information may be further obtained from the extracted landmark. For example, the age converter 60 may further acquire distance information between landmarks using Delaunay triangles.

The age conversion unit 60 generates a facial texture at the target age from the original face from which the landmark is extracted (that is, the face in step S410) (S420).

The texture at the target age is a facial texture expected when the subject reaches the target age. The facial texture is information representing the texture of the region of interest, and since the texture is unique for each facial component, it is a suitable factor to be used for age transformation. In an embodiment, the texture may be expressed as pattern data.

The age conversion unit 60 extracts the texture of the original face from which the landmark is extracted (that is, the texture at the original age) (S421), and applies the texture at the original age to the pre-learned texture transformation model. A face texture is generated at the age (S423).

In an embodiment, the texture at the original age may be a normal texture. A normal texture is a texture extracted from an image to which a separate preprocessing is not applied.

In another embodiment, the texture at the original age may be a shapeless face texture. The age converter 60 extracts the face shape of the original image, calculates the average shape of the face of the original image, and generates a face texture having the average shape as a shapeless face texture (S421). This is because the age conversion process for the shape is performed in the following step (S440).

In one embodiment, the texture transformation model includes a plurality of sub-models, each modeled to output a facial texture at a specific age. For example, when the device 1 is configured to process an age transformation to a preset target age (eg, to 40 years old) in response to an age transformation request, the texture transformation model is configured to output a face texture at the age of 40 can be configured. On the other hand, when the device 1 is configured to process the age transformation into a target age (eg, 25 years old, 60 years old, etc.) included in the request in response to the age transformation request, the texture transformation model is It contains a sub-model configured to output the face texture of In this case, the device 1 is further configured to provide the user with information about the age range that can be converted before receiving the age conversion request.

The texture transformation model (or sub-model) is a machine learning model capable of learning for one class while generating an image. The texture transformation model may be, for example, a Generative Adversarial Network (GAN)-based model as shown in FIG. 4 , but is not limited thereto.

For clarity of description, a texture transformation process will be described using a sub-model configured to transform a texture to a specific age having a GAN structure.

The sub-model having the GAN structure includes a generator and a discriminator. The generator is configured to output new data by applying noise to the input data. The generator aims to trick the discriminator into discriminating the similar data as real data by generating data similar to real data. The discriminator aims to identify the actual data and the output data of the generator.

As training progresses, the generator and discriminator update the parameters in the model to achieve their respective goals. The discriminator learns to reduce the probability of making a mistake, and the generator learns to increase the probability that the discriminator makes a mistake with respect to data output from random noise. That is, the generator and discriminator are trained to solve the aforementioned minimax problem.

In this learning process, the discriminator receives feedback from the correct answer of the input value (ie, training data), and the generator receives feedback from the discriminator. The process of learning such a GAN structure model is described in Non-Patent Document 1 (Goodfellow, Ian J.; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley, David; Ozair, Sherjil; Courville, Aaron; Bengio). , Yoshua (2014). "Conditional Generative Adversarial Networks"), a detailed description thereof will be omitted.

The texture transformation model is machine-learned using a plurality of texture training samples. Each texture training sample includes a facial texture of a training target at a specific age, respectively. In addition, each texture training sample may further include first label data indicating the age of the training target and second label data indicating the gender of the training target.

In an embodiment, the texture training sample includes at least a part of a shapeless facial texture obtained from a face of a training target having the specific age.

In an embodiment, when the texture transformation model includes a plurality of sub-models, the texture transformation model is machine-learned using a plurality of texture training sample sets. The texture training sample set for each sub-model includes a facial texture of a training target at a specific age corresponding to the sub-model. In some embodiments, each set of texture training samples includes at least a portion of shapeless facial textures obtained from the faces of a training subject having a corresponding age.

In addition, in some embodiments, each texture training sample may further include first label data indicating a specific age corresponding to a specific sub-model and second label data indicating the gender of the training target.

As described above, since each texture training sample set used for each sub-model includes information on a training target having the same specific age, each sub-model is learned through a training sample having the same specific age. Accordingly, each sub-model is associated with a corresponding specific age, and eventually the device 1 may search for a sub-model associated with the target age.

Meanwhile, each sub-model may be trained to be related to the age of an age group having a predetermined range. In this case, each training sample set used for each sub-model is learned using information of a training target belonging to the corresponding age group.

When the sub-model having the GAN structure is trained with a texture training sample set including a plurality of texture training samples at a specific age, at least some of the plurality of texture training samples may be used as real data.

Since the sufficiently learned and learned generator is configured to output data as close as possible to the input data, the pre-trained sub-model may output a texture at a target age. Here, the close output data is data in which a vector interval (eg, a feature vector) from a facial texture of a person who actually has a target age is minimized. The generator transforms input data to have a distribution of data at a specific age, and outputs the transformed data.

Accordingly, when the target age is a specific age, the transformed texture at the specific age output from the generator may be used as the texture at the target age.

In some embodiments, when a shape free face texture is used as a texture training sample, the generator outputs a transformed shapeless face texture close to the shape free face texture at a target age.

In another embodiment, the texture transformation model (or sub-model) may be a cycleGAN-based model.

cycleGAN is a model that is trained on unpaired data. CycleGAN includes two generators, and each generator outputs data converted to different domains. To this end, the noise of each generator may be different from each other. The discriminator of cycleGAN aims to identify the data of different domains output by each generator, respectively.

Since the submodel with the cycleGAN structure has two generators, it is further trained based on cycle consistency. The cyclic coherence is that when data converted from the first domain to the second domain is converted back to the first domain, it must be converted into an image previously generated in the first domain.

In this way, when using the cycleGAN-based texture transformation model, there is no need to prepare a paired data set, so it is easier to model the age transformation model of the face.

Since the training process of the cycleGAN-based texture transformation model (or sub-model) and the training sample used for learning are similar to the GAN-based texture transformation model (or sub-model), a detailed description will be omitted.

After step S421, the age conversion unit 60 applies the shapeless face texture as input data to the texture transformation model learned in advance for the target age to generate a shapeless face texture at the converted target age ( S423).

In an embodiment, the age conversion unit 60 receives a specific target age, and then performs pre-trained for the target age from among a plurality of sub-models each learned for a specific age included in the pre-trained texture transformation model. Search for submodels.

Then, the age conversion unit 60 applies the expressionless face texture of the original image from which the landmark is extracted to the searched sub-model as input data to generate a shapeless face texture at the target age (S423).

Meanwhile, in the above, the process of generating a facial texture at a target age has been described as an embodiment using a texture transformation model including a plurality of sub-models, but the process of generating a facial texture at a target age of the present invention is not limited thereto. does not

As such, in some embodiments of the present invention, the texture transformation model may be configured to include a plurality of GAN-based or cycleGAN-based sub-models that can be transformed into two domains.

Meanwhile, in another embodiment, the texture transformation model may be configured to output facial textures at a plurality of ages as a single model structure. In some embodiments, the plurality of ages may be ages corresponding to the plurality of clusters. The plurality of clusters may be age groups (eg, 20's, 30's, 40's, etc.). The texture conversion model having a single model structure may convert input data into facial textures for a plurality of age groups. In this case, the texture transformation model of the single model structure corresponds to a face texture at an age corresponding to the twenties (eg, 25 years old), a facial texture at an age corresponding to the thirties (eg 35 years old), and a 40's. It is possible to generate a facial texture and the like at the age (eg, 45 years of age).

The texture transformation model of the single model structure is learned through a plurality of training sample sets corresponding to each of a plurality of ages. Each training sample set may include a face image of a training target having a specific age, and first label data indicating the age of the training target and second label data indicating the gender of the training target. When the plurality of ages correspond to the plurality of clusters, each training sample set includes a face image of a training target belonging to the corresponding cluster.

In an embodiment, the texture transformation model of the single model structure is configured as a machine learning model that transforms input data into a plurality of domains, and is learned through an algorithm for this. For example, the texture transformation model of the single model structure may be a conditional GAN (cGAN)-based model. The cGAN-based model is similar to the GAN-based model, but is characterized in that information (y) indicating a specific condition is added to the generator and the discriminator. In the GAN architecture, the generator applies noise to transform it. In the cGAN architecture, the generator applies noise and information (y) to transform it. Here, the information y may be a class label indicating each of a plurality of ages. For example, it may be a class label representing each age group.

Since the process of learning the cGAN-based model is disclosed by Non-Patent Document 2 (Mehdi Mirza, Simon Osindero, (2014) “Generative Adversarial Nets”), a detailed description thereof will be omitted herein.

After step S421, the age conversion unit 60 applies the shapeless face texture as input data to the texture transformation model learned in advance for the target age to generate a shapeless face texture at the converted target age ( S423).

In one embodiment, the age conversion unit 60 applies the expressionless face texture of the original image from which the landmark is extracted in step S421 as input data to the texture conversion model of a single model structure. The texture transformation model of the single model structure generates facial textures at a plurality of ages. The age conversion unit 60 selects a facial texture at a target age from among facial textures at a plurality of ages (S423). To this end, the device 1 receives the target age in advance. The age conversion unit 60 uses a facial texture at an age matching the target age to generate an age conversion image.

Also, the age conversion unit 60 generates a face shape at the target age from the original face from which the landmark is extracted (S440).

To this end, the age converter 60 extracts the facial shape features of the original image based on the landmarks of the face (ie, the original face) of the original image (S431).

The face shape feature is a feature related to a face shape and is extracted using various feature extraction algorithms for extracting shape related features from an image. The feature extraction algorithm includes, but is not limited to, for example, PCA.

The extracted face shape feature may be expressed as various types of values. In an embodiment, the facial shape feature may be extracted as a feature vector. When N facial shape features are extracted, the feature vector is configured in N dimensions.

The age conversion unit 60 uses the facial shape features of the original image and a pre-learned shape transformation model to generate facial shape features at the requested age. The shape transformation model is pre-trained to output the facial shape features at the requested age when the facial shape features at the original age are applied.

The shape transformation model was created by modeling the relationship between age and facial shape features at that age.

In one embodiment, the shape transformation model is based on the difference between the age function value at the target age and the age function value at the original age and the facial shape feature of the original image (eg, the facial shape feature of step S431). modeled to calculate the facial shape features of the subject at the target age based on the

For example, the shape transformation model may be expressed by the following equation when N-dimensional facial shape features are extracted.

Here, i is a natural number less than or equal to N, where a _i ^new represents the i-th facial shape feature (eg, a feature vector value) at the target age, and a _i ^org is the i-th facial shape feature at the original age (eg, feature vector value), age _new is the target age, age _org is the original age, and f _i ^ap is the age function for the i-th facial shape feature.

5 is a diagram for explaining a process of determining an age function for a facial shape feature, according to an embodiment of the present invention.

Referring to FIG. 5 , when N-dimensional facial shape features are extracted, an age function for each facial shape feature is determined by modeling a relationship with the age of a training subject having the corresponding facial shape feature.

That is, the age functions for a plurality of face shape features are respectively learned using a plurality of shape training sample sets. Each shape training sample set includes a facial shape feature of the training target and a plurality of shape training samples at the corresponding age including the training target's age as label data.

Here, the age associated with the plurality of shape training sample sets includes various ages, such as a person having an age between 1 and 80 years as shown in FIG. 4 .

The facial shape feature of the training target is a facial shape feature obtained from a facial image of the training target at a corresponding age. In an embodiment, the facial shape feature is a feature vector related to a shape, and may include an N-dimensional feature vector (where N is an integer greater than or equal to 1).

The first shape training sample shown in FIG. 5 includes age information (3 years old in FIG. 5) of the first training target, and facial shape features ({a ₁ , a ₂ , a ₃ , ...a _M } of the first training target. ) is included. The second shape training sample includes age information (3 years old in FIG. 4 ) of the second training target and facial shape features ({a ₁ , a ₂ , a ₃ , ...a _M }) of the second training target.

The age function for each face shape feature is determined based on the distribution of age information of the training target for each shape feature. Therefore, the age function included in the shape transformation model depends on the dimension of the shape feature.

The age function for each shape feature vector is determined using various fitting algorithms that determine regularity from distributed information without information on regularity.

For example, the age function for the i-th shape feature (ie, the i-th age function) is approximated by a 3D polynomial fitting algorithm, but is not limited thereto. Each age function determined by the approximation is represented by a continuous line in a feature component-age graph plot, as shown in FIG. 5 .

As a result, the old function of the first shaped features (a _1), the first shaped features (a ₁₎ shown in Figure 5 from the age distribution of the plurality of training target to be determined. In addition, the old function of the three shape features (a _3), the third shape characteristics (a ₃₎ shown in Figure 5 from the age distribution of the plurality of training target to be determined.

As such, a total of m age functions are trained for the m feature components at the target age.

FIG. 6 is a diagram illustrating an age function for a second shape feature trained by the process of FIG. 5 .

By the process of generating the age function of FIG. 5 , an age function for the second shape feature a ₂ may be determined.

6 and Equation 1, when the age associated with the original image is 10 years old and the requested age is 80 years old, the age function value for the second shape feature at 80 years old is f ₂ ^ap (80), 10 The age function value for the second shape feature at age is f ₂ ^ap (10).

Then, the pre-trained shape transformation model is based on the age function (f ₂ ^ap _{) for the second shape feature of FIG. 6 and the second shape feature (a 2} ¹⁰ ) extracted from the face of the original image at the age of 80. An output value (a ₂ ⁸⁰ ) of the second shape feature may be calculated.

As such, the age conversion unit 60 may calculate the facial shape features (eg, first to Nth facial shape feature sets) at the target age by using the pre-learned shape transformation model ( S433 ).

The age converter 60 may restore the facial shape features at the target age output by the shape transformation model to generate the face shape at the target age.

For example, when the facial shape feature vector set at the target age is output, the facial shape at the target age is generated by restoring each facial shape feature vector value ( S435 ).

The age converter 60 may use various restoration algorithms to generate the face shape at the target age. The restoration algorithm may include, for example, a principal component analysis (PCA)-based restoration algorithm, but is not limited thereto.

The age conversion unit 60 warps the face texture at the target age to the shape of the face at the target age to generate the face at the target age as an age conversion face (S440).

In the above-described embodiments, the shape transformation model and the texture transformation model have been described only for the age transformation in a direction in which the target age is greater than the age of the object, but is not limited thereto.

Since the age function of the shape transformation model is configured to enable continuous and bi-directional age transformation, it is possible to obtain an age-transformed face shape in a direction in which the age of the subject is younger than the target age. In addition, when the sub-model for the target age and the target age is previously learned, the age-converted face texture in the young direction may also be obtained.

As described above, the background image is selected based on the user's input, or is automatically determined based on the original image and/or the age-converted image below.

The image synthesizing unit 70 synthesizes the face of the target at the target age with the face region of the background image, and generates a job image indicating a case in which the target has a corresponding job at the target age. That is, the image synthesizing unit 70 may generate a job image of the target based on a background image in which a face image and a job of the target at the target age are expressed, which are included in the age conversion image.

The face of the selected background image and the face at the target age may have differences in body structural characteristics such as size and skeleton, or differences in poses such as face angle and direction. In this case, when the face of the target age is synthesized as it is with the face of the background image, an unnatural job image is generated for the user. The image synthesizing unit 70 is configured to generate a natural job image by matching the face of the background image with the face at the target age.

In an embodiment, the image synthesizing unit 70 may extract a landmark from the face of the background image and/or the face at the target age. The occupational image generating apparatus 1 may further acquire location information and identification information of the landmark of the face of the background image, and location information and identification information of the landmark of the face at the target age through the image synthesizing unit 70 . . Here, the identification information includes anatomical facial features that each landmark means.

Since the landmark extraction operation of the image synthesizing unit 70 is performed similarly to the landmark extraction operation of the age conversion unit 60 , a detailed description thereof will be omitted.

In addition, the image synthesizing unit 70 is configured to perform various operations, such as mapping between landmarks, warping of a face shape, and transplantation of a synthesizing region, for image synthesizing.

The image synthesizing unit 70 is configured to map the landmark of the face of the age conversion image to the face region of the background image.

In an embodiment, the image synthesizing unit 70 may perform mapping on the face region of the background image based on anatomical facial features that the facial landmark shp_sim of the age conversion image means. The image synthesizing unit 70 maps the landmark of the background image indicating the same anatomical facial feature and the landmark of the face at the target age, respectively. For example, the landmark corresponding to the left end of the lips extracted from the face of the age conversion image (that is, the face of the target at the target age) is mapped to the landmark corresponding to the left end of the lips extracted from the face of the background image. .

In an embodiment, the image synthesizing unit 70 may map the landmark of the face of the age conversion image to the face region of the background image by at least partially minimizing the positions between the landmarks. The face region of the background image after mapping includes an existing landmark (shp_bg) set and a mapped landmark (shp_sim_t) set.

Due to this mapping operation, the positions of the facial components of the background image and the positions of the facial components of the age conversion image are matched, thereby minimizing the unnaturalness of synthesis between facial regions caused by differences in poses and the like.

In addition, the image synthesis unit 70 is configured to match the face of the age conversion image to the face of the background image.

In an embodiment, the image synthesizing unit 70 is configured to warp the facial texture of the age conversion image based on the mapped landmark (ie, the location of the mapped landmark shp_sim_t). For example, the image synthesizing unit 70 may warp the face texture of the age conversion image into a face shape based on the mapped landmark.

As described above with reference to FIG. 4 , since the facial texture feature is extracted based on the landmark, the facial texture is expressed in relation to the landmark.

The image synthesizing unit 70 warps the facial texture of the age transformation image constructed based on the facial landmark (shp_sim) of the age transformation image to the mapped landmark (shp_sim_t). Such warping is performed based on the relative position of the landmark mapped in the background image and the relative position of the landmark of the target's face in the age conversion image.

By the warping, the face texture and/or face shape of the background image and the face texture and/or face shape of the age conversion image are matched based on the mapped landmark ship_sim_t.

The image synthesizing unit 70 may match the face shape of the background image with the face shape of the age conversion image through various image editing techniques capable of bending and warping the image. The image editing technique includes, but is not limited to, passion image editing.

Due to this face shape matching, the unique texture characteristics and/or shape characteristics of the subject's face are matched with the unique texture characteristics and/or shape characteristics of the human face of the background image, so that areas other than the face of the background image The unnaturalness of the synthesis between (that is, the background region) and the face region of the age conversion image is minimized.

In addition, the image synthesizing unit 70 is further configured to warp the face texture of the background image to the location of the landmark of the face of the age conversion image mapped to the background image (ie, the mapped landmark (shp_sim_t)). For example, the image synthesizing unit 70 warps the face texture of the background image into a face shape based on the mapped landmark.

The image synthesizing unit 70 generates a synthesizing area mask based on a set of facial landmarks shp_bg_t of the warped background image.

The composite area mask is configured to filter data in the area inside the mask. The image synthesizing unit 70 applies the synthesized region mask to the warped age transformation image so that the face shape is matched to filter the face region of the target at the target age in which the shape is already matched, and the filtered face region of the target It can be implanted in the face region of the background image. The transplantation refers to any operation in which the background image is edited to include the filtered image, such as overlapping the image corresponding to the mask region in the background image, or replacing the image corresponding to the mask region in the background image.

Since the landmark (shp_bg_t) set and the landmark (shp_sim_t) set are the same, the composite area mask can filter the same area in the background image and the age conversion image as the composite area.

In an alternative embodiment, since the landmark (shp_bg_t) set and the landmark (shp_sim_t) set are the same, a composite area mask may be generated based on the landmark (ship_sim_t) set.

As a result of the transplantation, the image synthesizing unit 70 generates an occupational image.

In this way, the image synthesizing unit 70 may generate a job image by synthesizing the age-converted face region with the background image using the warped age-converted image and the synthesized region mask.

The above-described job image generating apparatus 1 may be applied to various systems configured to provide a job image supply service to a user. For example, the occupational image generating apparatus 1 is applied to a photobooth system or a kiosk system.

7 is a conceptual diagram of a photobooth system according to an embodiment of the present invention.

Referring to FIG. 7 , the photo booth system 1000 includes: a booth 100 at least partially surrounded by the job imaging apparatus 1 and a wall serving as a background to form a predetermined space in which an object can be located; includes In some embodiments, the photobooth system 1000 includes a front light 200 ; chair 300; And it may further include one or more of the shield 400.

In the system 1000, the occupational imaging apparatus 1 is a photo booth type photo-taking and occupational image output apparatus, and the apparatus 1 includes: a face photographing camera 9; computer equipment for image processing (eg, the age conversion unit 60 and/or the image synthesis unit 70); A front touch screen (ie, the input device 10 and the display device 40) is provided.

The user is positioned in the booth 100 where at least a part of the face is to be photographed by the device 1 . The booth 100 consists of a space in which one or more people can stand or sit on a chair.

The booth 100 provides a wall that can serve as a background. As shown in FIG. 7 , the booth 100 may be a partially open space, but is not limited thereto. In another embodiment, the booth 100 may be of a sealed type, in this case, the booth 100 may further include a door (not shown) for entering/exiting the user. The front light 200 is a device for properly illuminating a user as a subject in the booth 100 during photo taking. The chair 300 is configured so that one or more people can sit on it. For example, a child as an object and a guardian protecting the child may sit on the chair 300 .

When the booth 100 is of an open type, the shield 400 is configured to at least partially block the open portion of the booth 100 in order to at least partially reduce the effect of light from outside the booth 100 on acquisition of the original image. do. As shown in FIG. 7 , the shield 400 may be configured in a curtain type, but is not limited thereto.

It will be apparent to those skilled in the art that the occupational image generating apparatus 1 and a system including the same may include other components not described herein. It may also include other hardware elements necessary for the operations described herein, including, for example, storage for storing data or information. In addition, the apparatus 1 or a system (eg, a photobooth) including the apparatus 1 may further include a network, a network interface and a protocol, and the like.

A method of generating a job image having an age-converted face according to another aspect of the present invention is provided to a computing device including a processor (eg, the job image generating device 1) and a system including the same (eg, photo booth 1000). can be performed by

Hereinafter, for clarity of explanation, a 7-year-old child is assumed and the target age is 25 years old, and the occupational image generation method will be described in more detail. However, it will be apparent to those skilled in the art that the present invention is not to be understood as being limited to the age conversion of a 7 year old child to 25 years old.

8 is a flowchart of a method for generating a job image, according to an embodiment of the present invention.

Referring to FIG. 8 , the method of generating a job image having an age-converted face includes: obtaining an original image including a face of a subject at a specific age (S10); obtaining original age information of the subject (S15); determining a background image in which a job is expressed, to be used for generating a job image (S20); obtaining target age information to be converted (S25); converting the face of the subject at the original age into the face of the subject at the target age (S30); and generating a job image of the target by synthesizing the age-converted face with the face region of the background image (S50). In some embodiments, the method includes: Career guidance including at least a job image among the original image of step S10, the background image of step S20, the age conversion image of step S30, and the job image of step S50 The method may further include providing a service image to the user (S60). In addition, the method includes: after generating the job image (S50), requesting confirmation of the user's satisfaction with the job image (S51); and when an input indicating dissatisfaction is received, returning to the previous step (eg, S10, S15, S20, S25, S30 or S50) and performing the corresponding step and subsequent steps again (S55). .

In steps S10, S15, S20, and S25, image data and detailed information used to generate an age-converted image for generating an occupational image are obtained. For example, original image data including the face of the 7-year-old child is acquired in step S10, information about the age of 7 corresponding to the age of the subject is acquired in step S15, and in step S20 A background image expressing a job desired by a 7-year-old child is acquired, and information about a 25-year-old corresponding to the target age is acquired in step S25.

In step S10, the original image data is acquired by the transceiver 20 by the photographing device 9 of the device 1 or from the user's mobile device.

In one embodiment, the information on the original age in step S15 is obtained as a user input (eg, input by the input device 10 or the transceiver 20 of the device 1).

In another embodiment, the information on the original age in step (S15) is obtained by automatically analyzing the original image.

In addition, in step S15, gender information on the subject of the original image may be further obtained. Gender information on the subject of the original image is obtained by the above-described user input or analysis from the original image.

The background image in step S20 may be determined by a user input, or may be automatically determined based on the original image of step S10 and/or the age-converted image of step S30. In some embodiments, when the background image is determined based on the age conversion image, step S20 is performed after step S30.

In one example, the background image representing the job is selected based on a user input. To this end, the device 1 or the computing device for performing the method may be configured to provide a job selection screen and an interface for receiving an input. For example, the device 1 may provide the user with a selection menu including various job titles such as scientist, police officer, firefighter, doctor, and singer.

The device 1 receives a job selection command for the selection menu through an input device, and in response to it, retrieves a background image related to the selected job (S20). The input for selecting the background image may be received together with detailed information related to the original image in step S20 or may be received through an interface for selecting the background image. Then, the matched background image is used to generate a job image (S50).

The target age in step S25 may be determined by a user input or may be automatically obtained according to the determination of the background image in step S20. For example, in step S25 , a value preset to be related to the background image determined in step S20 (eg, the average age of the occupation of the background image) may be obtained as the target age.

In step S30, the face of the target of the original image is converted into the face of the target at the target age to generate the age-converted face of the target. For example, the face of a 7-year-old child is transformed into an adult's face at 25 years old to generate an age-converted face expected to have when the child turns 25 years old.

In an embodiment, the converting into the age-converted face (ie, the target's face at the target age) includes: extracting a landmark from the target's face of the original image; generating a facial texture of the target at the target age from the target's face of the original image from which the landmark is extracted; generating a face shape of the subject at the target age from the face of the subject of the original image from which the landmark is extracted; and generating an age-converted face of the subject based on the face texture and face shape of the subject at the target age.

In an embodiment, the step of extracting the landmark from the face of the target of the original image may be performed through a preset landmark extraction algorithm. In some embodiments, before extracting the landmark, the step of detecting the face region of the target from the original image may be performed first.

In an embodiment, the generating of the facial texture of the target at the target age includes: generating a shapeless facial texture from the target's face (eg, face image) of the original image from which the landmark is extracted; and applying the shapeless facial texture to a pre-learned texture transformation model to generate a shapeless facial texture of the target at the target age.

In an embodiment, the shapeless face texture may be data obtained by converting the face shape of the target of the original image into an average shape.

In one embodiment, the texture transformation model is pre-trained to output a facial texture at a target age. For example, the texture transformation model is pre-trained to output the face texture at the age of 25.

In an embodiment, the texture transformation model may be a GAN-based model. Here, the texture transformation model includes a generator trained in advance to output a transformation texture corresponding to the facial texture at the target age by applying noise to the input data.

The generator was pre-trained using a transform texture and a discriminator that identifies reference data (ie real data) similar to the transform texture.

Since the generator outputs a texture converted to be very similar to the facial texture at the actual target age, the transformed texture output from the generator may be used as a facial texture of the target at the target age. That is, the generator is configured to output data corresponding to the facial texture at the target age.

The texture transformation model (eg, generator) is generated using a plurality of training samples, each training sample including a facial texture of a training target at the age of 25.

In some embodiments, the texture transformation model may include a pre-trained sub-model for the target age to output a facial texture at the target age. For example, the texture transformation model includes a pre-trained first sub-model to output a face texture at the age of 25. In this case, the sub-model is generated using a training sample set consisting of a plurality of training samples, and each set represents the face texture of the training target at the corresponding age, first label data indicating the corresponding age, and the gender of the training target. and second label data. The texture transformation model was pre-trained using a plurality of training sample sets depending on the number of sub-models.

In another embodiment, the texture transformation model may be a cycleGAN-based model. Here, the texture transformation model includes a plurality of generators (eg, two) for generating a facial texture of the target at the target age. The plurality of generators are configured to output data converted into different domains. The plurality of generators were previously trained using a discriminator configured to identify data of different domains output by each generator.

For example, the two generators may include: a first generator that applies noise to input data of a first domain to output converted data of a second domain; and a second generator for outputting transformed data of a third domain by applying noise to the input data of the first domain, wherein each generator is configured to re-convert the transformed data into data in the first domain to satisfy cyclic consistency. and is configured to be converted into input data of the first domain.

In another embodiment, the pre-trained texture transformation model may be a texture transformation model having a single model structure that generates facial textures at a plurality of ages from single input data. In this case, the facial texture at the target age is selected from facial textures at a plurality of ages.

In some embodiments, the plurality of ages may be ages corresponding to the plurality of clusters. For example, the plurality of clusters may be age groups (20's, 30's, 40's, etc.).

The texture conversion model of the single model structure generates face textures at a plurality of ages by applying noise and condition information (y) to input data. Here, the condition information y includes a class label indicating each of a plurality of ages (eg, a class label indicating an age for each age group).

The texture transformation model of the single model structure may be, for example, a conditional GAN-based model, but is not limited thereto.

In an embodiment, the generating of the face shape of the subject at the target age includes: extracting facial shape features of the subject of the original image based on the landmark of the face of the original image; generating facial features of the target at the target age by applying the facial shape features of the original image to a pre-learned shape transformation model; and reconstructing the facial shape of the target at the target age based on the facial shape features of the target at the target age.

The shape transformation model is generated by modeling the relationship between age and the facial shape features of the training target at the corresponding age, and the difference between the age function value at the target age and the age function value at the original age and the original image It is modeled based on the facial shape features (eg, the facial shape features of step S431). The shape transformation model may be expressed, for example, by Equation 1 above.

The shape transformation model may be modeled based on an age function for each facial shape feature when the dimension of the facial shape feature is N-dimensional (where N is an integer greater than or equal to 1).

The shape transformation model is a model pre-trained using a plurality of training samples and label data indicating the target age to output facial shape features at the target age, and the training sample in each set is a training target at the corresponding age. of facial features. For example, the Moang transformation model is trained in advance so as to output the facial shape features at the age of 25 using the facial shapes of a plurality of 25-year-old training targets.

In some embodiments, when the shape transformation model is configured to output a facial shape feature at any one of a plurality of ages, the shape transformation model includes a plurality of training samples to output a facial shape feature for each of the plurality of ages. A model pre-trained using a set, wherein each set includes a plurality of training samples at a specific one of the plurality of ages and label data representing the specific age, and a training sample in each set is a training target at the corresponding age. of facial features.

In one embodiment, the generating of the age-converted face of the subject includes: warping the face texture at the target age with the face shape of the subject at the target age, and transforming the face at the target age into age It may include generating as a face.

9 is a diagram exemplarily showing a result according to an age conversion operation, according to an embodiment of the present invention.

Referring to FIG. 9 , an age-converted image in which an original image including the subject's face as a child at the age of 7 is converted into the face of the subject as an adult at the age of 25 may be generated.

After step S30, the age-converted face is synthesized in the face region of the background image determined in step S20 to generate a job image of the target (S50).

10 is a flowchart of an image synthesis process according to an embodiment of the present invention.

Referring to FIG. 10 , the image synthesis process is: extracting the landmark of the age-converted face of the target in step S30, and extracting the landmark of the face of the background image in step S20 (S510). The landmark extraction of step S510 is performed through the same or similar process as the landmark extraction of step S410.

11 is a diagram illustrating a landmark extraction result of an age conversion image and a background image according to an embodiment of the present invention.

Referring to FIG. 11 , a job selected by the user to generate a job image is a police officer, and a background image is an image representing a police officer as a job. In step S510, a landmark set including a plurality of landmarks is extracted from the age-converted face of the subject and the face of the background image, respectively.

In some embodiments, the landmark of the face of the background image may be pre-stored together with the background image. In this case, a pre-stored landmark related to the searched background image may be used in step S510 when the background image is searched.

Based on each extracted landmark, the landmark of the age-converted face is mapped to the face region of the background image (S520).

12 is a diagram illustrating a landmark of an age conversion image mapped to a background image, according to an embodiment of the present invention.

Referring to FIG. 12 , the landmarks of the age conversion image shown in FIG. 11 are mapped to the background image.

In one embodiment, the mapping of step S520 is performed based on anatomical facial features that each landmark means.

After mapping, the position of the landmark of the face of the age conversion image is moved to the position of the landmark mapped to the background image (S530). As a result, the face shape of the age conversion image may be warped to match the face shape of the background image.

13 is a diagram illustrating a result of moving a landmark of a face of a background image according to an embodiment of the present invention.

Referring to FIG. 13 , the landmark of the age conversion image shown in FIG. 11 moves to the mapping point of FIG. 12 . In step S530, the face shape of the age conversion image may be warped based on the moved landmark position of the age conversion image.

Also, after mapping, the position of the landmark of the face of the background image is moved to the position of the landmark mapped to the background image (S540).

14 is a diagram illustrating a result of positional movement of a landmark of a face of an age conversion image according to an embodiment of the present invention.

Referring to FIG. 14 , the landmark of the background image shown in FIG. 11 moves to the mapping point of FIG. 12 ( S540 ). In step S540, the face shape of the background image may be warped based on the moved landmark position of the background image.

Due to the execution of steps S530 and S540, the face shape of the age conversion image is matched with the face shape of the background image.

The position movement of step S530 or S540 is performed based on the mapping result (ie, mapping point) of step S520 .

Then, based on the landmark of the face of the background image, a composite area mask for filtering the inner area may be generated (S550).

In an embodiment, the composite area mask is generated by using a landmark located at the outermost part of the moved landmark of the background image.

15 is a diagram illustrating a composite area mask according to an embodiment of the present invention.

By using the moved landmark of the face of the background image of FIG. 13 , the composite area mask shown in FIG. 15 can be obtained.

After the step (S550), the face region of the target of the warped age transformation image is filtered using the synthesized region mask (S560). Then, a job image is generated by transplanting the filtered target face region to the face region of the background image (S560).

Due to the warping according to the movement of the landmark position in steps S530 and S540, the face shape of the target of the age conversion image matches the face shape with the background image. The face region of the target at the target age in which the face shape is matched with the background image is filtered through the composite region mask.

16 is a diagram illustrating a composite image when a target job is a police officer according to an embodiment of the present invention.

Referring to FIG. 16 , it can be confirmed that the face shape of the age conversion image of FIG. 9 is warped by the age conversion process. In step S560, the face region of the subject of the warped age conversion image is transplanted into the background image representing the police.

As a result, a job image representing a case in which a 7-year-old object becomes a police officer at the age of 25 may be generated (S560).

17 is a diagram illustrating a composite image when a target job is a doctor, according to an embodiment of the present invention.

Referring to FIG. 17 , it can be seen that the face shape of the age conversion image of FIG. 9 is warped somewhat differently from FIG. 16 . This is because the mapping result ( S520 ) using the background image representing the doctor is different from the mapping result ( S520 ) using the background image representing the police.

In step S560, the face region of the target of the warped age conversion image is transplanted into the background image expressing the intention. As a result, a job image representing a case in which a 7-year-old object becomes a police officer at the age of 25 may be generated (S560).

The above-described job image generation process is not limited to using a still image. In one embodiment, the job image for the background image may be a moving picture consisting of a plurality of frames. In this case, at least one of the plurality of frames includes an image (eg, a still background image) representing a job.

In this case, at least one frame of the still age conversion image and the background video is used as a still background image to generate a job image for the frame.

The result of synthesizing the age conversion image and the background image may be generated as a moving image (hereinafter, “job moving image”) consisting of a plurality of frames. In an embodiment, the job video is generated based on a video in which a still age conversion image and a background image appear (hereinafter, “background video”).

For example, by using at least one frame of the still age conversion image and the background video as a still background image, a job image for the frame is generated, and a job video composed of a frame having the job image is generated.

Since the process of generating the job image for each frame is similar to the process of generating the job image based on the still age conversion image and the still background image, a detailed description will be omitted.

The result of generating the job image in step S50 is provided to the user (S60).

In step S60, the final result including a combination including at least a composite photo among an input photo (ie, an original image), a background photo (ie, a background image), and a composite photo (ie, a job image) is generated as a career guidance service image and a screen displaying the career guidance service image is provided to the user.

A frame, sticker, text, picture, etc. that can be customized by the user may be inserted into the career guidance service image.

18A and 18B are diagrams illustrating a career guidance service image according to embodiments of the present invention.

Referring to FIG. 18A , the career guidance service image provided to the user may include an input photo, a composite photo, a background border, and a logo (KIST) of a company that manufactures the device 1 or provides a career guidance service. .

Referring to FIG. 18B , the career guidance service image may include a background picture, a composite picture, a background border, and a company logo (KIST).

Such a career guidance service image may be provided to the user through photo printing, communication such as e-mail or mobile phone transmission, or display by a display device. The device 1 and a system including the same (eg, a photo booth) may provide a career guidance service image to assist a subject in determining their own career path.

In an alternative embodiment, immediately after generating the job image, before providing the career guidance service image, after generating the job image (S50), the user's satisfaction with the job image is checked (S51).

In step S51, if the currently generated job image is provided to the user through photo printing, communication such as e-mail, mobile phone transmission, or display by a display device, the device 1 is satisfied with the provided current job image. is further configured to confirm For example, “Are you satisfied with the video? - Yes/No” or “Do you want to print (or print) the image? - A screen that requests input of judgment on satisfaction/dissatisfaction, such as “yes/no”, may be provided in the form of a whole or a pop-up.

If an input corresponding to satisfaction is received in step S51, the device 1 provides a final job image based on the background image determined in step S20 through step S60.

If an input corresponding to dissatisfaction is received in step S51, the device 1 performs the steps from at least one of the previous steps (S10, S15, S20, S25, S30 and S50) to the job image generation (S50). The processes are performed again (S55).

In an embodiment, the device 1 is further configured to induce, after receiving an input corresponding to a dissatisfaction, an input for determining a starting stage of re-progression. The device 1 provides the user with a list of dissatisfaction indicating the items of dissatisfaction for each step (S10, S15, S20, S25, S30, S50), and when the user inputs at least one item of dissatisfaction, A re-run may be performed from the step associated with the corresponding item. For example, the dissatisfaction list may include “dissatisfied with the original image”, “dissatisfied with the conversion start age”, “dissatisfied with the background image”, “dissatisfied with the target age”, “dissatisfied with the age conversion image”, and “dissatisfied with the job image” , when an input for “dissatisfied with the background image” is received, the device 1 performs the operations from step S20 to step S50 again.

After that, if a job image is generated according to the re-progress, dissatisfaction is re-confirmed again (S51), and when an input corresponding to satisfaction is received, a career guidance service image based on the re-generated job image is provided to the user ( S60).

The apparatus for generating an occupational image 1 according to the embodiments described above, a system including the same, and an operation by the method for generating an occupational image may be at least partially implemented as a computer program and recorded in a computer-readable recording medium. . For example, embodied with a program product consisting of a computer-readable medium containing program code, which may be executed by a processor for performing any or all steps, operations, or processes described.

The computer may be any device that may be incorporated into or may be a computing device such as a desktop computer, laptop computer, notebook, smart phone, or the like. A computer is a device having one or more alternative and special purpose processors, memory, storage, and networking components (either wireless or wired). The computer may run, for example, an operating system compatible with Microsoft's Windows, Apple OS X or iOS, a Linux distribution, or an operating system such as Google's Android OS.

The computer-readable recording medium includes all types of recording identification devices in which computer-readable data is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage identification device, and the like. In addition, the computer-readable recording medium may be distributed in a network-connected computer system, and the computer-readable code may be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present embodiment may be easily understood by those skilled in the art to which the present embodiment belongs.

Although the present invention as described above has been described with reference to the embodiments shown in the drawings, it will be understood that these are merely exemplary and that various modifications and variations of the embodiments are possible therefrom by those of ordinary skill in the art. However, such modifications should be considered to be within the technical protection scope of the present invention. Accordingly, the true technical protection scope of the present invention should be defined by the technical spirit of the appended claims.

Occupational images generated by the embodiments of the present invention may be utilized in a way of providing an image service after job experience at a ^{child job experience specialized facility (eg, KidZania TM} , Job World ^{TM ).} In addition, it can be utilized as a method of providing content such as educational videos introducing various occupations by synthesizing the adult face of the age-converted child. In addition, it can be used as a souvenir for birthdays, graduations, children's day, etc. based on the future hopes of children surveyed in early childhood education facilities such as daycare centers and kindergartens.

In particular, by generating an age-converted image at a target age using a model based on machine learning, which is one of the quaternary industrial fields, it is possible to provide a natural job image showing the future appearance of infants.

Therefore, since interest in children's jobs is increasing in the age of low fertility, industrial application is expected to be very high.

Claims (20)

In the photo booth,
The photo booth includes a device for generating a job image having an age-converted face, including a processor; and a booth at least partially surrounded by a wall to form a space where the object can be located,
The device is:
obtaining an original image including the subject's face at the original age;
obtaining original age information of the subject;
determining a background image in which a job is expressed for use in generating a job image;
converting the subject's face at the original age into the subject's face at the target age; and
and synthesizing the age-converted face with the determined face region of the background image to generate a job image of the target,
The step of generating the job image of the target is,
extracting each landmark of the face and age-converted face of the background image;
mapping the landmark of the age-converted face to the face region of the background image based on each extracted landmark;
In order to warp the face texture of the age transformation image having the age transformation face, the landmark of the age transformation face having a position on the age transformation image having the age transformation face is mapped to the background image. moving it to a position on the job image of
moving the landmark of the face of the background image having the position on the background image to the position on the job image of the landmark of the age-converted face mapped to the background image in order to warp the face texture of the background image;
generating a composite area mask for filtering an inner area based on the moved landmark of the face of the background image; and
and filtering the face region of the warped age transformation image using the composite region mask, and transplanting the filtered face region to the face region of the background image.
According to claim 1,
Photobooth, characterized in that it is further configured to perform the step of providing a career guidance service image including at least a job image among the original image, the background image, the age conversion image, and the job image to the user.
According to claim 1, wherein the step of determining the background image for the job image,
providing an interface screen for inducing a user's input for selecting a background image by displaying at least some of the previously stored candidate background images; and
When an input for selecting a background image is received, determining a candidate background image corresponding to the input as a background image for the job image of the target;
The candidate background image is configured to express the characteristics of the job and is different from the target photo booth, characterized in that it includes at least a part of a face of a person having a corresponding job.
According to claim 3, The step of determining the background image for the job image,
Prior to providing the interface screen for inducing a user's input to select the background image, further comprising the step of providing a selection screen for selecting a desired job by displaying a job-related list,
The job-related list includes at least one of one or more job listings and one or more job group entries.
The method of claim 3, wherein the interface screen,
When the subject's gender information is obtained, the photo booth comprising at least a part of the candidate background image related to the subject's gender.
The method of claim 1 , wherein the converting into the face of the subject at the target age comprises:
extracting a landmark from the face of the target of the original image;
generating a facial texture of the target at the target age from the target's face of the original image from which the landmark is extracted;
generating a face shape of the subject at the target age from the face of the subject of the original image from which the landmark is extracted; and
Photo booth comprising the step of generating an age-converted face of the target based on the facial texture and face shape of the target at the target age.
The method of claim 6, wherein the generating of a facial texture at the target age comprises:
generating a shapeless face texture from the face of the target of the original image from which the landmark is extracted; and
and applying the shapeless facial texture to a pre-learned texture transformation model to generate a shapeless facial texture of the target at the target age.
The method of claim 7, wherein the texture transformation model,
a first generator for outputting transformed data of a second domain by applying noise to input data of a first domain; and
A second generator for outputting converted data of a third domain by applying noise to the input data of the first domain,
and each generator is configured to be converted into input data of the first domain upon re-conversion of the converted data into data into the first domain.
The method of claim 8, wherein the texture transformation model,
A photobooth comprising: a generator that applies noise and condition information to input data of a first domain to output converted data of a plurality of different domains including a second domain and a third domain.
The method of claim 8, wherein the texture transformation model,
A model trained in advance using a plurality of training samples to output data corresponding to a facial texture at the target age, wherein each training sample is a label including a facial texture of a training target at the target age, and a gender of the training target A photobooth comprising data.
8. The method of claim 7, wherein generating the face shape of the subject at the target age comprises:
extracting facial features of the target of the original image based on the landmark of the face of the original image;
generating facial features of the target at the target age by applying the facial shape features of the original image to a pre-learned shape transformation model; and
and restoring a facial shape of the target at the target age based on the facial shape features of the target at the target age.
The method of claim 11, wherein the shape transformation model,
generated by modeling the relationship between age and facial features at that age,
It is characterized in that the model is modeled to calculate the facial shape features of the target at the target age based on the difference between the age function value at the target age and the age function value at the original age and the facial shape features of the target in the original image. photo booth.
The method of claim 12, wherein the shape transformation model,
When the dimension of the facial feature is N-dimensional (where N is an integer greater than or equal to 1), the photobooth is characterized in that it is modeled based on an age function for each facial feature.
The method of claim 11, wherein the shape transformation model,
A model pre-trained using a plurality of training samples and label data indicating the target age to output facial shape features at the target age, wherein the training samples in each set include facial shape features of a training target at the corresponding age Photo booth, characterized in that.
The method of claim 7, wherein the generating of the age-converted face of the target comprises:
and warping a face texture at the target age with the face shape of the target at the target age, and generating the face at the target age as an age-converted face.
delete The method of claim 1, wherein the mapping comprises:
Photobooth, characterized in that it is performed based on the anatomical facial features that each landmark means.
The method of claim 2, wherein to provide the career guidance service image,
The photo booth, characterized in that at least one of displaying the career guidance service image, printing the career guidance service image, and transmitting the career guidance service image.
delete According to claim 1,
lighting providing light to the object; a chair on which the subject sits; A photo booth further comprising at least one of a screen to block the outside and the inside of the booth.

Images