Privacy-Preserving Action Recognition via Motion Difference Quantization

TL;DR

This paper introduces BDQ, a privacy-preserving encoder for human action recognition that effectively balances action recognition accuracy with privacy protection by filtering sensitive information through a three-stage process.

Contribution

The paper proposes a novel end-to-end trainable encoder, BDQ, that combines blurring, motion difference, and quantization to preserve privacy while maintaining action recognition performance.

Findings

Achieves state-of-the-art privacy-accuracy trade-off on benchmark datasets.

Performs comparably to DVS sensor-based event cameras in privacy preservation.

Demonstrates robustness across multiple datasets.

Abstract

The widespread use of smart computer vision systems in our personal spaces has led to an increased consciousness about the privacy and security risks that these systems pose. On the one hand, we want these systems to assist in our daily lives by understanding their surroundings, but on the other hand, we want them to do so without capturing any sensitive information. Towards this direction, this paper proposes a simple, yet robust privacy-preserving encoder called BDQ for the task of privacy-preserving human action recognition that is composed of three modules: Blur, Difference, and Quantization. First, the input scene is passed to the Blur module to smoothen the edges. This is followed by the Difference module to apply a pixel-wise intensity subtraction between consecutive frames to highlight motion features and suppress obvious high-level privacy attributes. Finally, the Quantization module is applied to the motion difference frames to remove the low-level privacy attributes. The BDQ parameters are optimized in an end-to-end fashion via adversarial training such that it learns to allow action recognition attributes while inhibiting privacy attributes. Our experiments on three benchmark datasets show that the proposed encoder design can achieve state-of-the-art trade-off when compared with previous works. Furthermore, we show that the trade-off achieved is at par with the DVS sensor-based event cameras. Code available at: https://github.com/suakaw/BDQ_PrivacyAR.