摘要 Abstract 

人臉去識別化旨在隱藏個人的身分資訊，同時保留臉部影像中可支援後續應用所需的重要資訊。本演講將回顧人臉去識別化技術的發展歷程，探討不同方法如何在隱私、實用性與視覺品質之間取得平衡。此外，本演講也將介紹近年逐漸受到重視的研究方向，聚焦於如何發展更具實用性、更穩健，並能以系統化方式進行評估的隱私保護的電腦視覺系統。

Face de-identification aims to conceal a person’s identity while preserving the facial information needed for useful downstream applications. This talk surveys the evolution of face de-identification, examining how different approaches balance privacy, utility, and visual quality. It also discusses emerging directions toward more useful, robust, and systematically evaluated privacy-preserving vision systems.