Method and system for generating image including specific object

The method leverages a LoRA model to efficiently generate images with specific faces and backgrounds by pre-training on specific datasets and adjusting the base model for face replacement, addressing the resource constraints and performance issues of current AI models in generating specific images.

WO2024248378A1PCT designated stage expired Publication Date: 2024-12-05LINE PLUS
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
PCT/KR2024/006758
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-31
Filing Date
2024-05-20
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Current text-to-image AI models, such as stable diffusion, require significant resources and time for training and are not optimized for generating specific images, particularly faces and objects, leading to poor performance when generating full-body images with specific faces or proper backgrounds.

Method used

A method using a Low Rank Adaptation (LoRA) model, which is smaller and faster to train, is employed to generate images by pre-training on specific datasets, and then adjusting a base model to replace the face in a person image with a specific person's face using image-to-image conversion, ensuring the final image resembles the training data.

Benefits of technology

This approach allows for efficient generation of images with specific faces and backgrounds, improving the quality of full-body images by focusing on the face replacement process, reducing the need for extensive training datasets and resource-intensive learning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure KR2024006758_05122024_PF_FP_ABST
    Figure KR2024006758_05122024_PF_FP_ABST
Patent Text Reader

Abstract

A method and a system for generating an image including a specific object are disclosed. The method for generating an image, according to one embodiment, may comprise the steps of: receiving a designated base model pre-trained to generate an image; receiving a designated prompt and parameter; generating a person image through text-to-image conversion according to the prompt and the parameter by using the base model; generating a masking image obtained by masking the face of the person image; using the base model and an image set of a specific person so as to receive a designated additional model pre-trained to generate an image of the specific person; and using the person image and the masking image as inputs to generate a final image for the specific person through image-to-image conversion based on the base model adjusted through the additional model.
Need to check novelty before this filing date? Find Prior Art

Description

Method and system for generating an image containing a specific object

[0001] The description below relates to a method and system for generating an image containing a specific object.

[0002] When creating a text-to-image AI model, such as stable diffusion, the AI ​​model is trained by inputting a large number of images and / or photographs along with descriptive text about those images and / or photographs. After training, the generated AI model file is passed to an inference tool, and prompts and parameters are specified, generating images described by the prompts. The prompts describe the desired image, which can be entered in natural language, such as "a cat jumping over a roof." Parameters can also include settings for noise methods and counts.

[0003] However, the training process of these AI models itself requires significant costs (e.g., large amounts of GPUs (Graphics Processing Units) and time). Furthermore, because the training dataset is trained to generate all types of images, including people, animals, objects, and backgrounds, it does not generate specific faces of individuals (e.g., celebrities) or specific objects (e.g., specific brands of beverages). Furthermore, because the images generated vary depending on the training data, many models specialize in generating specific images. For example, a model that trains heavily on portraits will be good at generating portraits, while a model that trains heavily on illustrations will be good at generating illustrations.

[0004] The LoRA (Low Rank Adaptation) model is a compact model trained on an additional dataset based on a specific stable diffusion model. LoRA can be an additional learning technique that trains a new subject to an existing model. These LoRA models are smaller in size, allowing them to train faster and take up less memory than stable diffusion models. However, their small size prevents them from training on a wide variety of data. A LoRA model can be created by training an image dataset (images from various angles / poses) of an object, such as a specific person, using a specific stable diffusion model as the base model. In other words, a LoRA model trained on an image dataset of an object, such as a specific person, can be generated based on a pre-trained stable diffusion model. Meanwhile, when using a LoRA model, a file for a specific stable diffusion model can be specified in an inference tool, and additionally, a specific LoRA model can be specified, such as prompts and parameters. This can generate images similar to the LoRA model's training data. In other words, images with a specified prompt feel can be obtained while being close to the dataset used to train the LoRA model.

[0005] However, when training a LoRA model using a dataset containing a specific person's face, the model successfully generates images containing the face. However, when generating a full-body image containing the person's face, the model fails to properly generate the face. This problem arises because the LoRA model is trained to reproduce the dataset as is. Similarly, because the LoRA model attempts to generate the entire image as if it were a face, it suffers from the problem of not properly depicting elements such as the background.

[0006] A method and system are provided for generating an image including the face of a characteristic person by changing the face of a person image generated using a base model into the face of a specific person.

[0007] A method for generating an image in a computer device including at least one processor, the method comprising: designating a base model pre-trained to generate an image by the at least one processor; designating a prompt and parameters by the at least one processor; generating a person image by performing text-to-image conversion according to the prompt and the parameters using the base model by the at least one processor; generating a masking image by masking a face of the person image by the at least one processor; designating an additional model pre-trained to generate an image of the specific person by using the base model and an image set of the specific person; and generating a final image of the specific person by performing image-to-image conversion based on the base model adjusted through the additional model, using the person image and the masking image as inputs, by the at least one processor.

[0008] According to one aspect, the step of generating the final image may be characterized by generating a final image of the specific person by deforming the person image and replacing the masked face portion in the masking image with the face of the specific person learned through the additional model.

[0009] According to another aspect, the step of generating the final image may be characterized by including a step of adjusting the base model so that the base model is learned for the specific person using the additional model.

[0010] According to another aspect, the base model may include a stable diffusion model, and the additional model may be characterized in that it is generated through stable diffusion learning using the base model and an image set of the specific person.

[0011] According to another aspect, the additional model may be characterized by including a LoRA (Low Rank Adaptation) model for learning the specific person in the base model.

[0012] According to another aspect, the step of generating the final image may be characterized by generating a final image for the specific person through image-to-image conversion according to the prompt and the parameters.

[0013] A computer program stored on a computer-readable recording medium is provided for executing the above method on a computer device in combination with a computer device.

[0014] A computer-readable recording medium having recorded thereon a program for executing the above method on a computer device is provided.

[0015] A computer device is provided, comprising at least one processor implemented to execute computer-readable instructions, wherein the computer device is configured to receive a pre-trained base model for generating an image by the at least one processor, receive a prompt and a parameter, generate a person image through text-to-image conversion according to the prompt and the parameter using the base model, generate a masking image in which a face of the person image is masked, and receive a pre-trained additional model for generating an image of the specific person using the base model and an image set of the specific person, and generate a final image of the specific person through image-to-image conversion based on the base model adjusted through the additional model, using the person image and the masking image as inputs.

[0016] An image containing the face of a featured person can be created by changing the face of a person image created using a base model to the face of a specific person.

[0017] FIG. 1 is a block diagram illustrating an example of a computer device according to one embodiment of the present invention.

[0018] FIG. 2 is a diagram illustrating an example of a process for generating (learning) a LoRA model according to an embodiment of the present invention.

[0019] FIG. 3 is a diagram illustrating an example of an image generation process using a LoRA model according to one embodiment of the present invention.

[0020] FIG. 4 is a diagram illustrating examples of actual images generated using a LoRA model trained with an actual photo dataset, according to one embodiment of the present invention.

[0021] Figure 5 is a diagram showing examples of real images generated using a pre-trained stable diffusion model and prompts as a comparative example.

[0022] FIGS. 6 to 9 are drawings illustrating examples of an image generation process for improving a method of generating a face portion when using a LoRA model in one embodiment of the present invention.

[0023] Figure 10 is a flowchart illustrating an example of an image generation method according to one embodiment of the present invention.

[0024] Hereinafter, the detailed description will be given with reference to the attached drawings.

[0025] An image generation system according to embodiments of the present invention may be implemented by at least one computer device. In this case, a computer program according to an embodiment of the present invention may be installed and run on the computer device, and the computer device may perform an image generation method according to embodiments of the present invention under the control of the run computer program. The above-described computer program may be stored on a computer-readable recording medium, which may be coupled to the computer device and cause the computer to execute the image generation method.

[0026] FIG. 1 is a block diagram illustrating an example of a computer device according to an embodiment of the present invention. As illustrated in FIG. 1, the computer device (100) may include a memory (110), a processor (120), a communication interface (130), and an input / output interface (I / O interface) 140. The memory (110) is a computer-readable recording medium and may include a random access memory (RAM), a read only memory (ROM), and a permanent mass storage device such as a disk drive. Here, the ROM and the permanent mass storage device such as the disk drive may be included in the computer device (100) as a separate permanent storage device distinct from the memory (110). In addition, an operating system and at least one program code may be stored in the memory (110). These software components may be loaded into the memory (110) from a computer-readable recording medium separate from the memory (110). This separate computer-readable recording medium may include a computer-readable recording medium such as a floppy drive, a disk, a tape, a DVD / CD-ROM drive, a memory card, etc. In another embodiment, the software components may be loaded into the memory (110) via a communication interface (130) other than a computer-readable recording medium. For example, the software components may be loaded into the memory (110) of the computer device (100) based on a computer program installed by files received via a network (Network, 160).

[0027] The processor (120) may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to the processor (120) via the memory (110) or the communication interface (130). For example, the processor (120) may be configured to execute instructions received according to program code stored in a storage device such as the memory (110).

[0028] The communication interface (130) may provide a function for the computer device (100) to communicate with other devices via a network (160). For example, requests, commands, data, files, etc. generated by the processor (120) of the computer device (100) according to program codes stored in a recording device such as a memory (110) may be transmitted to other devices via the network (160) under the control of the communication interface (130). Conversely, signals, commands, data, files, etc. from other devices may be received by the computer device (100) via the communication interface (130) of the computer device (100) via the network (160). Signals, commands, data, etc. received via the communication interface (130) may be transmitted to the processor (120) or the memory (110), and files, etc. may be stored in a storage medium (the aforementioned permanent storage device) that the computer device (100) may further include.

[0029] The input / output interface (140) may be a means for interfacing with an input / output device (I / O device, 150). For example, the input device may include a device such as a microphone, a keyboard, or a mouse, and the output device may include a device such as a display or a speaker. As another example, the input / output interface (140) may be a means for interfacing with a device that integrates input and output functions, such as a touchscreen. The input / output device (150) may also be configured as a single device with the computer device (100).

[0030] Additionally, in other embodiments, the computer device (100) may include fewer or more components than those illustrated in FIG. 1. However, it is not necessary to explicitly illustrate most conventional components. For example, the computer device (100) may be implemented to include at least some of the input / output devices (150) described above, or may further include other components such as a transceiver, a database, etc.

[0031] FIG. 2 is a diagram illustrating an example of a process for generating (learning) a LoRA model according to an embodiment of the present invention. FIG. 2 shows an example in which a pre-trained stable diffusion model A (210) is designated as a base model. At this time, it is assumed that the stable diffusion model A (210) has been pre-trained to generate a photographic image. As training data, a photographic dataset (220) of a specific object (e.g., a specific person) can be used. It is assumed that the photographic dataset (220) includes images of various angles and / or poses for the specific object. The embodiment of FIG. 2 shows that a LoRA model X (240) can be generated through a process of stable diffusion learning (230) using the stable diffusion model A (210) designated as a base model and the photographic dataset (220). Since the technology itself for stable diffusion learning for generating a LoRA model is already well known, a detailed description thereof will be omitted. In this embodiment, we focus on creating a LoRA model X (240) using a pre-trained stable diffusion model A (210) to generate a photo image and a photo dataset (220) for a specific object.

[0032] FIG. 3 is a diagram illustrating an example of an image generation process using a LoRA model according to an embodiment of the present invention. FIG. 3 shows an embodiment in which a pre-trained stable diffusion model A (210) to generate a photographic image is used as a base model in a stable diffusion text-to-image conversion (310). At this time, the stable diffusion text-to-image conversion (310) can generate an image (340) by specifying a file of the stable diffusion model A (210) to an inference tool, and specifying a prompt (320), a parameter (330), and a LoRA model X (240) generated through the process of FIG. 2.

[0033] FIG. 4 is a diagram illustrating examples of actual images generated using a LoRA model trained with an actual photo dataset according to an embodiment of the present invention. The embodiment of FIG. 4 shows an example of a LoRA model (420) generated according to the learning method described in the embodiment of FIG. 2 above using an actual photo dataset (410) of a specific person A. Here, the LoRA model (420) of FIG. 4 may be a model trained to generate a photo image of a specific person A.

[0034] At this time, when an image for a specific person A is generated using a LoRA model (420) trained with an actual photo dataset (410) for a specific person A, it can be seen that images for the specific person A are generated well, as in the first result (430). However, when an image for a specific person A is generated using the LoRA model (420) by additionally specifying clothing / motion / background, etc. as a prompt (in the embodiment of FIG. 4, “full body, outdoors, walking, city background,” etc. are specified), it can be seen that desired images are not generated well, as in the second result (430). That is, when the proportion of the face in the entire image generated using the LoRA model (420) is large, the image is generated well, but when the proportion of the face is small, the desired images are not generated well.

[0035] Fig. 5 is a comparative example, showing examples of actual images generated using a pre-trained stable diffusion model and prompts. The comparative example of Fig. 5 shows a third result (510) generated using the pre-trained stable diffusion model A (210) described in Figs. 2 and 3 and prompts without using the LoRA model. In this case, the prompts were identically specified as "full body, outdoors, walking, city background." In this case, unlike the second result (430), it can be seen that the third result (510) successfully generated an image of a person walking with his or her whole body in a city background.

[0036] This problem can be solved by inserting learning images with various backgrounds, poses, and compositions when training the LoRA model (420), but obtaining learning images with various backgrounds, poses, and compositions for a specific person incurs a large cost, and even if these learning images are obtained, the training itself incurs a large cost.

[0037] FIGS. 6 to 9 are drawings illustrating examples of an image generation process for improving a method of generating a face portion when using a LoRA model in one embodiment of the present invention.

[0038] In the embodiment of FIG. 6, an example of a process of first specifying a file of a stable diffusion model (610) to an inference tool, and then specifying a prompt (620) and parameters (630) to generate a person image (650) based on the prompt through stable diffusion text-to-image conversion (640) is shown. At this time, the image (710) of FIG. 7 shows an example of an actually generated person image (650). In this case, the prompt (620) is properly applied to the person image (650), but it includes a person other than the specific person learned by the LoRA model (670).

[0039] Hereinafter, the embodiment of FIG. 6 shows an example of a process of manually / automatically detecting a face in a generated person image (650) and adding a mask to the detected face to generate a masked image (660). At this time, the image (810) of FIG. 8 shows an example of adding a mask (820) to the face area of ​​the person in the image (710) of FIG. 7. For example, the mask (820) may be generated by overlaying the area of ​​the detected face part with a specific color. The addition of the mask (820) may be performed by manually finding the area of ​​the face, but may also be performed by automatically detecting it using a well-known technology such as Haar Cascades of OpenCV.

[0040] Hereafter, in the embodiment of FIG. 6, the stable diffusion model (610) can be designated as the base model. In addition, similarly to the embodiment of FIG. 2, the stable diffusion model (610) can be adjusted so that a specific person is learned in the stable diffusion model (610) through a LoRA model (670) learned for a specific person. The person image (650) can be used as an image-to-image input, and the masked image (660) can be used as an image-to-image inpaint mark input. Inpaint can be a function of the stable diffusion for changing a specific part of the image. In the present embodiment, a function for changing the masked part (face part) into the face of a specific person learned through the LoRA model (670) can be provided. In this case, the image-to-image input and the image-to-image inpaint marks input can be input into the adjusted stable diffusion model (610) to generate the final image (690) through the stable diffusion image-to-image transformation (680). Depending on the embodiment, the prompt (620) and parameter (630) described above may be further used as inputs to the stable diffusion image-to-image transformation (680).

[0041] This final image (690) may be an image in which the face of a person (650) has been replaced with the face of a specific person learned by the LoRA model (670). The face replacement may be achieved by deforming only the face portion through a masked image (660). The image (910) of FIG. 9 illustrates an example of an actually generated final image (690).

[0042] Through this process, we can generate images of various poses, clothing, backgrounds, etc. of a specific person using only a facial dataset of that person, without having to have multiple datasets for that person.

[0043] Fig. 10 is a flowchart illustrating an example of an image generation method according to an embodiment of the present invention. The image generation method according to the present embodiment can be performed by a computer device (100). At this time, the processor (120) of the computer device (100) can be implemented to execute control instructions according to the code of the operating system included in the memory (110) or the code of at least one computer program. Here, the processor (120) can control the computer device (100) so that the computer device (100) performs steps (1010 to 1060) included in the method of Fig. 10 according to the control instructions provided by the code stored in the computer device (100).

[0044] In step (1010), the computer device (100) may be assigned a pre-trained base model to generate an image. For example, the base model may include, but is not limited to, a stable diffusion model. For example, the base model may be any artificial intelligence model capable of generating images through text-to-image conversion based on natural language input, such as a prompt, and capable of learning about a specific person through an additional model described below.

[0045] At step (1020), the computer device (100) may be assigned prompts and parameters. The prompt may be a description of the image to be generated, which may be entered in natural language, such as "a cat jumping over a roof." Furthermore, the parameters may include settings for noise methods, counts, etc.

[0046] In step (1030), the computer device (100) can generate a person image through text-to-image conversion according to prompts and parameters using a base model. Here, the person image may be an image that includes a person other than the specific person ultimately desired to be generated, according to the description of the prompt.

[0047] In step (1040), the computer device (100) may generate a masking image that masks the face of the person image. For example, the computer device (100) may detect a face area of ​​the person image and then generate a masking image by adding a mask to the detected face area. As another example, the computer device (100) may also receive a masking image in which a mask is added to the face of the person image.

[0048] In step (1050), the computer device (100) may be assigned an additional model pre-trained to generate an image of a specific person using a base model and an image set of a specific person. The additional model may be trained using an image set reflecting various angles and / or poses of a specific person to generate a photographic image of a specific object based on the base model. This additional model may include, but is not limited to, a LoRA model for training the base model with a specific person. For example, a model generated to generate an image of a specific person by training a pre-trained base model with a specific person to generate an image may be utilized as the additional model.

[0049] In step (1060), the computer device (100) may generate a final image of a specific person through image-to-image transformation based on a base model adjusted through an additional model, which inputs a person image and a masking image. At this time, the computer device (100) may adjust the base model using the additional model so that the base model is learned for the specific person. Thereafter, the computer device (100) may transform the person image based on the adjusted base model, and replace the masked face portion in the masking image with the face of the specific person learned through the additional model to generate a final image of the specific person. In some embodiments, the computer device (100) may also generate a final image of the specific person through image-to-image transformation according to the prompts and parameters specified in step (1020).

[0050] In this way, according to embodiments of the present invention, an image including the face of a characteristic person can be created by changing the face of a person image created using a base model into the face of a specific person.

[0051] The systems or devices described above may be implemented as hardware components, or a combination of hardware components and software components. For example, the devices and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0052] Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or may independently or collectively command the processing device. The software and / or data may be embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0053] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., singly or in combination. The medium may continuously store a computer-executable program, or may temporarily store it for execution or download. In addition, the medium may be various recording means or storage means in the form of a single or multiple hardware combinations, and is not limited to a medium directly connected to a computer system, but may also be distributed over a network. Examples of the medium may include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and those configured to store program commands, including ROM, RAM, and flash memory. In addition, examples of other media may include recording media or storage media managed by app stores that distribute applications, sites that supply or distribute various software, servers, etc. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0054] Although the embodiments described above have been described by way of limited examples and drawings, those skilled in the art will appreciate that various modifications and variations can be made based on the above teachings. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0055] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

1. A computer program stored in a computer-readable recording medium for executing an image generation method on a computer device by being combined with a computer device, The above image generation method is, A step of specifying a pre-trained base model to generate images; Steps to specify prompts and parameters; A step of generating a human image through text-to-image conversion according to the prompt and the parameters using the base model; A step of creating a masking image by masking the face of the above person image; A step of designating an additional model pre-trained to generate an image of a specific person using the base model and an image set of the specific person; and A step of generating a final image of the specific person through image-to-image transformation based on the base model adjusted through the additional model, using the person image and the masking image as inputs. including A computer program characterized by .

2. In paragraph 1, The steps for generating the final image are: Transforming the above-mentioned person image, and replacing the masked face part in the masking image with the face of a specific person learned through the additional model to generate a final image of the specific person. A computer program characterized by .

3. In paragraph 1, The steps for generating the final image are: A step of adjusting the base model so that the base model is learned for the specific person using the additional model. A computer program comprising:

4. In paragraph 1, The above base model includes a stable diffusion model, The above additional model is created through stable diffusion learning using the base model and the image set of the specific person. A computer program characterized by .

5. In paragraph 1, The above additional model includes a LoRA (Low Rank Adaptation) model for learning the specific person in the base model. A computer program characterized by .

6. In paragraph 1, The steps for generating the final image are: Generating a final image of the specific person through image-to-image transformation according to the above prompt and parameters. A computer program characterized by .

7. A method for generating an image of a computer device including at least one processor, A step of specifying a pre-trained base model to generate an image by at least one processor; A step of specifying a prompt and parameters by at least one processor; A step of generating a human image through text-to-image conversion according to the prompt and the parameters using the base model by at least one processor; A step of generating a masking image by masking a face of the person image by at least one processor; A step of designating an additional model pre-trained to generate an image of a specific person using the base model and an image set of the specific person by at least one processor; and A step of generating a final image of the specific person through image-to-image transformation based on the base model adjusted through the additional model, using the person image and the masking image as inputs by at least one processor. A method for generating an image including:

8. In paragraph 7, The steps for generating the final image are: An image generation method characterized in that the above-mentioned person image is transformed, and a final image of the specific person is generated by replacing the masked face portion in the masking image with the face of the specific person learned through the additional model.

9. In paragraph 7, The steps for generating the final image are: A step of adjusting the base model so that the base model is learned for the specific person using the additional model. An image generation method characterized by including:

10. In paragraph 7, The above base model includes a stable diffusion model, The above additional model is created through stable diffusion learning using the base model and the image set of the specific person. A method for generating an image characterized by .

11. In paragraph 7, The above additional model includes a LoRA (Low Rank Adaptation) model for learning the specific person in the base model. A method for generating an image characterized by .

12. In paragraph 7, The steps for generating the final image are: An image generation method characterized in that a final image of the specific person is generated through image-to-image transformation according to the above prompt and the above parameters.

13. A computer-readable recording medium having recorded thereon a computer program for executing the method of any one of claims 7 to 12 on a computer device.

14. At least one processor implemented to execute computer-readable instructions. Including, By at least one processor, We are given a pre-trained base model to generate images, Given prompts and parameters, Generate a human image through text-to-image conversion according to the above prompt and the above parameters using the above base model, Create a masking image by masking the face of the above person image, An additional model pre-trained to generate an image of a specific person using the base model and an image set of the specific person is specified, Generating a final image of the specific person through image-to-image transformation based on the base model adjusted through the additional model, using the person image and the masking image as inputs. A computer device characterized by:

15. In paragraph 14, To generate the final image, by the at least one processor, Transforming the above-mentioned person image, and replacing the masked face part in the masking image with the face of a specific person learned through the additional model to generate a final image of the specific person. A computer device characterized by:

Citation Information

Patent Citations

  • Thin film transistor and method of manufacturing the same

    KR1020220137863A

  • Airbag cushion Protector

    KR1020240157415A

  • Evaluating System for Capability of Biometric Terminal to Detect Fake Biometric and Evaluating Method Thereof

    KR1020240165110A