Invention Title:

VIRTUAL MAKEUP SOLUTION PROVIDING SYSTEM USING GENERATIVE ARTIFICIAL INTELLIGENCE

Publication number:

US20260120368

Publication date:

2026-04-30

Section:

Physics

Class:

G06T11/60

Inventors:

JongChul YE 🇰🇷 Daejeon, South Korea

Myeongjin Goh 🇰🇷 Yongin-si, South Korea

Seongmin Jeong 🇰🇷 Yongin-si, South Korea

Geonyeong PARK 🇰🇷 Daejeon, South Korea

Serin YANG 🇰🇷 Daejeon, South Korea

Hee Chan JEON 🇰🇷 Yongin-si, South Korea

Inhwa HAN 🇰🇷 Daejeon, South Korea

Assignees:

Korea Advanced Institute of Science and Technology 🇰🇷 Daejeon, South Korea

AMOREPACIFIC CORPORATION 🇰🇷 Seoul, South Korea

Applicants:

KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY 🇰🇷 Daejeon, South Korea

AMOREPACIFIC CORPORATION 🇰🇷 Seoul, South Korea

Smart overview of the Invention

The system leverages generative artificial intelligence (AI) to offer a virtual makeup solution. It comprises a user terminal that captures a face image and a makeup service providing server. The server processes the image to create a virtual makeup effect, using a combination of feature extraction and color application technologies. The system is designed to provide a seamless and realistic virtual makeup experience.

Technical Details

The server includes several components: a receiving part for face image acquisition, a feature extraction model for analyzing facial features, and a color extraction part that utilizes a Large Language Model (LLM)-based generative AI. This AI extracts makeup colors corresponding to the facial features. The final step involves a diffusion-based generative AI that applies the extracted colors to the face image, ensuring a natural appearance.

Background and Innovation

In recent years, the beauty industry has increasingly incorporated AI technologies to offer personalized services. Previous methods often lacked the ability to consider detailed facial features or resulted in a mismatch between the virtual and real appearances. The proposed system addresses these issues by considering both facial shape and color, ensuring a more accurate and harmonious virtual makeup application.

System Components

The feature extraction model identifies facial contours and components using advanced techniques like BiseNet and ResNet. It assesses various facial attributes such as skin tone, lip color, and iris color. The LLM-based AI is trained with datasets to accurately associate facial features with suitable makeup colors. The diffusion-based AI, specifically Stable Diffusion, generates the final makeup image by integrating the selected colors into predefined facial areas.

Benefits and Applications

This virtual makeup system not only enhances the user experience by providing realistic makeup previews but also aids in making informed purchasing decisions. By simulating actual makeup application, users can explore and choose optimal makeup colors and products tailored to their individual features, thus streamlining the selection process and potentially boosting sales in the beauty industry.