KR-20260067596-A - Virtual fitting service method using face-specific creation model

Abstract

The present invention relates to a virtual fitting service method using a face-specialized generative model, characterized by comprising the steps of: receiving a clothing image selected from a user terminal and user body shape information; a preprocessing module preprocessing a body shape image corresponding to the selected clothing image and the user body shape information; a face-specialized generative model generating a latent vector for generating a face region among the body shape images; and a clothing synthesis model generating a clothing synthesis result by synthesizing the selected clothing image onto the body shape image using the preprocessing result and the latent vector. According to the present invention as described above, distortion of facial features during clothing synthesis can be prevented, thereby improving the quality of the virtual fitting result and maximizing the effect of virtual fitting.

Inventors

• 김민선

Assignees

• (주)하마랩

Dates

Publication Date

20260513

Application Date

20241106

Claims (4)

1. A step of receiving a selected clothing image from a user terminal and user body shape information; A preprocessing module preprocesses the selected clothing image and the body shape image corresponding to the user body shape information; A step in which a face generation specialized model generates a latent vector for generating a face region among the above body shape images; A virtual fitting service method using a face-specific generative model, characterized by including the step of generating a clothing synthesis result in which the clothing synthesis model synthesizes the selected clothing image onto the body shape image using the preprocessing result and the latent vector.
2. In paragraph 1 A virtual fitting service method using a face-specialized generative model, characterized in that the above clothing synthesis model encodes feature information of the clothing image, combines feature information of the body shape image, the latent vector and the feature information of the clothing image, and passes it through a self-focusing network and a mutual focus network to generate the clothing synthesis result.
3. In paragraph 1, The above clothing synthesis model uses a Stable Diffusion model, and A virtual fitting service method using a face-specialized generative model, characterized by preparing a pre-trained basic synthetic model and an untrained additional synthetic model, measuring the difference in output between the two models for the same input through loss, updating the weights of the additional synthetic model through backpropagation, and using the additional synthetic model with updated weights as the clothing synthetic model.
4. In paragraph 3, A virtual fitting service method using a face-specialized generative model, characterized in that the clothing synthesis model calculates the difference between the outputs of two models for the same input as KL diversity loss, and calculates the difference between the output of the additional synthesis model and the label value as cross-entropy loss.

Description

Virtual fitting service method using face-specific creation model The present invention relates to a virtual fitting service method using generative artificial intelligence, and more specifically, to a technology that prevents distortion of facial features during clothing synthesis by performing facial region learning and clothing synthesis learning through separate models using a face-specific generative model. The reason consumers prefer offline shopping over online shopping when buying clothes is the strong perception that clothing is a product that must be tried on in person to ensure a good fit. Due to this consumer preference for trying on clothes before purchasing, online clothing sales have not grown as actively, and this purchasing trend is even more pronounced for expensive items. Meanwhile, consumers often search for and refer to reviews, such as purchase reviews and blog reviews, to try and alleviate the inconvenience of offline shopping. However, since these reviews are merely accounts of others' experiences, they are often not of much help to the consumer themselves. Consequently, there are many cases where consumers end up making a final purchase only after visiting an offline store in person and trying on the clothes they wanted. As a result, there is a growing demand for virtual fitting services that allow users to check whether clothing fits their body type when purchasing online. As a technology for such virtual fitting services, Korean Registered Patent No. 2580427 presents a technology that provides an environment where users can have an experience similar to trying on clothes in person offline by providing a virtual fitting image through the synthesis of the clothing image onto the model image when users upload only clothing images and model images. However, conventional virtual fitting service technologies suffered from the problem of distorted model facial images during the process of synthesizing clothing images and model images using generative AI, which resulted in a degradation of the quality of the virtual fitting results. In addition, conventional virtual fitting services have the problem that they are not suitable for providing real-time services online because the clothing synthesis process requires a significant amount of time. This patent is the result of a project filed with the support of Gyeonggi Province and the Gyeonggi Economic & Science Promotion Agency's 2024 Global Startup Commercialization Support Program. FIG. 1 shows the configuration of a virtual fitting system using a face-specific generative model according to the present invention. Figure 2 is a block diagram illustrating the internal configuration of a virtual fitting service device. Figure 3 is a block diagram illustrating the detailed configuration of a virtual fitting model. FIG. 4 is a flowchart illustrating the process of performing a virtual fitting service using a face-specific generative model according to the present invention. The embodiments described in the present invention and the configurations illustrated in the drawings are merely preferred embodiments of the present invention and do not represent all of the technical concept of the present invention; therefore, the scope of the rights of the present invention should not be interpreted as being limited by the embodiments and drawings described in the text. That is, since the embodiments are subject to various modifications and may take various forms, the scope of the rights of the present invention should be understood to include equivalents capable of realizing the technical concept. Furthermore, the objectives or effects presented in the present invention do not imply that a specific embodiment must include all of them or only such effects; therefore, the scope of the rights of the present invention should not be understood as being limited by them. Unless otherwise defined, all terms used herein have the same meaning as generally understood by those skilled in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having meanings consistent with the context of the relevant technology and should not be interpreted as having an ideal or overly formal meaning not explicitly defined in this invention. Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to the attached drawings. FIG. 1 shows the configuration of a virtual fitting system using a face-specific generative model according to the present invention, FIG. 2 is a block diagram showing the internal configuration of a virtual fitting service device, and FIG. 3 is a block diagram showing the detailed configuration of a virtual fitting model. As illustrated in FIG. 1, a virtual fitting system using a face-specific generative model according to the present invention is configured to include a user terminal (100) and a virtual fitting service device (200). The user terminal (100) is