ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper Into the Portraits: Face Swapping System for Art Gallery Visitors
Cited 0 time in scopus Share share facebook twitter linkedin kakaostory
Authors
Juwon Lee, Jung-Jae Yu, Seong Jae Yoo, Wonyoung Yoo
Issue Date
2023-10
Citation
International Symposium on Mixed and Augmented Reality (ISMAR) 2023, pp.718-721
Publisher
IEEE
Language
English
Type
Conference Paper
DOI
https://dx.doi.org/10.1109/ISMAR-Adjunct60411.2023.00155
Abstract
To provide a unique and personalized experience for visitors at the art gallery, we have developed a face swapping system that seamlessly replaces the original face in a painting with the face of visitor. This system aims to create a sense of immersion, making the visitor feel as if the artist had originally depicted their portrait and they had become part of the artwork itself. In order to achieve this, we constructed a dataset of portrait paintings, including manual annotation of facial landmarks for a selected subset, which serves as the target images for our system. We developed a photo- to-paint face swapping method by building upon the FaceShifter and fine-tuning it using the constructed dataset. To ensure the high-quality results, we have incorporated several post-processing steps into the pipeline. To make this system accessible to gallery visitors, we implemented a photo kiosk based on the developed method. This kiosk has been successfully installed at Théâtre des Lumières, a digital art center located in Seoul, South Korea. Since its installation, the kiosk has gained significant popularity, with around 20,000 visitors utilizing it and generating a substantial number of 31,820 images. This high level of visitor engagement demonstrates the widespread appeal and popularity of the system.
KSP Keywords
Art Gallery, Digital art, Facial landmarks, High-quality, Manual annotation, Post-Processing, Sense of immersion, South Korea, Visitor engagement, face swapping, fine-tuning