딥러닝 기반 사용자의 가상 패션 착용 이미지 및 숏폼 생성 프로그램 개발 = Development of a Deep Learning-based Program for User’s Virtual Try-on Images and Short Forms Generation in Fashion

다국어 초록 (Multilingual Abstract)

The study presents the development of an immersive virtual fitting program leveraging deepfake technology and its extension to video formats. Traditional online shopping environments have been limited by challenges such as the inability to try on clothing, lack of realism, and restricted selection of apparel, thereby constraining the consumer experience. This research integrates advanced deep learning technologies, including StableVITON, DeepFaceLab, and MusePose, to seamlessly synthesize consumer facial images with model images and adapt selected garments appropriately. Furthermore, motion synthesis was implemented to provide virtual fitting results in video format, enabling a more realistic and natural user experience. The results demonstrate that the program achieves high-quality synthesis not only for apparel from specific retailers but also for garments from diverse sources, offering a highly realistic fitting experience. Future developments will focus on optimizing code efficiency and expanding into 3D domains to deliver faster and more precise virtual fitting services.