Privacy

Responsible and ethical AI: Pedestrian privacy protection

Team

  • Dr. Xingchen Zhang, Senior Lecturer, University of Exeter. Former Marie-Curie Individual Fellow at Imperial College London
  • Mr. Zixian Zhao, PhD student, University of Exeter
  • Prof. Yiannis Demiris, Professor, Imperial College London

Introduction

Many videos are captured to train AI models. We aim to protect pedestrian privacy in videos captured by cameras mounted on robots and vehicles while maintaining the utility of the anonymized videos.

Traditional anonymization methods

First Image Second Image

However, traditional anonymization methods will affect the utility of the anonymized data for downstream tasks.

Pedestrian anonymization via face swapping

First Image Second Image

Pedestrian anonymization via full-body anonymization

This work is under review.

First Image

Related publications

  1. X. Zhang*, Z. Zhao. More effort is needed to protect pedestrian privacy in the era of AI. NeurIPS Position Paper Track, 2025. Oral paper. [PDF]
  2. Z. Zhao, X. Zhang*, Y. Demiris. 3PFS: Protecting pedestrian privacy through face swapping, IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 11, pp. 16845-16854, 2024.
  3. Z. Zhao, X. Zhang*, Y. Demiris. PedPrivacy: Pedestrian Privacy Protection through Pose Transferring and Face Swapping. Under review.

Funding

  • The Royal Society Research Grant, PI
  • Marie Sklodowska-Curie Individual Fellowship, PI
First Image