Visual Privacy

Visual privacy research focuses on mitigating the risks of sensitive information leakage from images and videos processed by computer vision systems. Current efforts concentrate on developing and auditing privacy-preserving techniques, including those based on differential privacy, diffusion models, and lensless imaging, often employing instruction tuning or self-supervised learning to enhance model performance while safeguarding privacy. This field is crucial for responsible AI development, addressing ethical concerns and legal implications surrounding the use of visual data in applications ranging from surveillance to activity recognition and impacting the design and deployment of privacy-aware computer vision systems.