PrivHAR: Recognizing Human Actions From Privacy-preserving Lens
This work addresses privacy concerns in action recognition applications, offering a solution for users and developers in surveillance and security domains, though it appears incremental as it builds on existing privacy-preserving methods.
The paper tackled the problem of balancing privacy protection with action recognition in video by proposing a framework that degrades video quality to inhibit privacy attributes and adversarial attacks while preserving features for activity recognition, validated through simulations and hardware experiments.
The accelerated use of digital cameras prompts an increasing concern about privacy and security, particularly in applications such as action recognition. In this paper, we propose an optimizing framework to provide robust visual privacy protection along the human action recognition pipeline. Our framework parameterizes the camera lens to successfully degrade the quality of the videos to inhibit privacy attributes and protect against adversarial attacks while maintaining relevant features for activity recognition. We validate our approach with extensive simulations and hardware experiments.