DSAIL Lab (지도교수: 한진영), CIKM Short paper 채택
- 인공지능융합학과(일반대학원)
- 조회수1906
- 2023-08-14
DSAIL Lab (지도교수: 한진영)의 정주호(인공지능융합학과), 강채원(인공지능융합학과), 윤지우(인공지능융합학과) 학생들이 연구한 논문 “SAFE: Sequential Attentive Face Embedding with Contrastive Learning for Deepfake Video Detection”이 세계 최고 권위 정보검색(Information Retrieval) 및 데이터마이닝 학회인 CIKM 2023 (32nd ACM International Conference on Information and Knowledge Management), Short papers에 채택되었습니다. 논문은 23년 10월 영국 버밍엄에서 발표될 예정입니다.
본 논문은 효과적인 딥페이크 비디오 탐지를 위해 딥페이크 비디오 영상 속에서 얼굴의 동적인 특징을 잡아낼 수 있는 SAFE (Sequential Attentive Face Embedding) 모델을 제안하였습니다. 이전의 연구들과 달리, 이 모델은 얼굴에서 나타나는 지역 정보(Local dynamics)와 전체 정보(global dynamics) 모두를 고려하여 비디오 영상의 진위여부를 파악하고, 나아가 contrastive learning을 통해 학습 과정을 최적화하였습니다. 논문의 자세한 내용은 다음과 같습니다.
[논문] Juho Jung, Chaewon Kang, Jeewoo Yoon, Simon S. Woo, Jinyoung Han, “SAFE: Sequential Attentive Face Embedding with Contrastive Learning for Deepfake Video Detection,” In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (CIKM 2023), Oct. 2023.
[Abstract].
The emergence of hyper-realistic deepfake videos has raised significant concerns regarding their potential misuse. However, prior research on deepfake detection has primarily focused on image based approaches, with little emphasis on video-based detection. With the advancement of generation techniques enabling intricate and dynamic manipulation of entire faces as well as specific facial components in a video sequence, capturing dynamic changes in both global and local facial features is crucial in detecting deepfake videos. This paper proposes a novel sequential attentive face embedding, SAFE, that can capture facial dynamics in a deepfake video. The proposed SAFE can effectively integrate global and local dynamics of facial features revealed in a video sequence using contrastive learning. Through a comprehensive comparison with the state-of-the-art methods on the DFDC (Deepfake Detection Challenge) dataset and the FaceForensic++ benchmark, we show that our model achieves the highest accuracy in detecting deepfake videos on both datasets.