RAPrivacy: a Readable Anonymizer for Privacy Preserving Action Recognition


ZiZhen Wang (National Yang Ming Chiao Tung University), Yen-Lung Chu (National Yang Ming Chiao Tung University), Pei-Chun Tsai (National Yang Ming Chiao Tung University), Kuan-Wen Chen (National Yang Ming Chiao Tung University)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

The growing awareness and demand for privacy protection have become increasingly prominent in the field of action recognition, which involves processing image sequences that capture human motion. Consequently, ensuring both accuracy and protecting sensitive information presents a significant challenge. Existing methods often face difficulties in balancing privacy preservation with effective action recognition, often resulting in images that are not perceptible to the human eye. To address these, we introduce RAPrivacy, a novel model that leverages adversarial learning in conjunction with style transfer to generate images that maintain privacy while remaining interpretable for human observers in the context of action recognition. For example, such images enable medical personnel to remotely monitor a patient’s physical condition without revealing the patient's entire identity. Empirical evaluations on the VP-UCF101 and VP-HMDB51 datasets demonstrate the effectiveness of RAPrivacy, achieving competitive performance in action recognition while significantly enhancing privacy protection. Notably, to the best of our knowledge, this is the first study in the domain of privacy-preserving action recognition that ensures human eye readability.

Citation

@inproceedings{Wang_2025_BMVC,
author    = {ZiZhen Wang and Yen-Lung Chu and Pei-Chun Tsai and Kuan-Wen Chen},
title     = {RAPrivacy: a Readable Anonymizer for Privacy Preserving Action Recognition},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_117/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection