Visual analysis tasks, including crowd management, often require resource-intensive machine learning models, posing challenges for deployment on edge hardware. Consequently, cloud computing emerges as a prevalent solution. To address privacy concerns associated with offloading video data to remote cloud platforms, we present a novel approach using adversarial training to develop a lightweight obfuscator neural network. Our method focuses on pedestrian detection as an example of visual analysis, allowing the transformation of video frames on the camera itself to retain only essential information for pedestrian detection while preserving privacy. Importantly, the obfuscated data remains compatible with publicly available object detectors, requiring no modifications or significant loss in accuracy. Additionally, our technique overcomes the common limitation of relying on labeled sensitive attributes for privacy preservation. By demonstrating the inability of pedestrian attribute recognition models to detect attributes in obfuscated videos, we validate the efficacy of our privacy protection method. Our results suggest that this scalable approach holds promise for enabling camera usage in video analytics while upholding personal privacy.

In this paper, we propose a novel annotation placement approach for revealing information about foreground objects in surveillance videos. To arrange positions of annotations, spatiotemporal coherence between annotations and foreground objects is applied. The annotation placement problem is formulated as an optimization problem with respect to spatiotemporal coherence of annotations and foreground objects. The optimization problem is effectively solved using Markov random fields. To the best of our knowledge, this paper is the first work that discusses and solves the annotation placement problem for surveillance videos by considering the relationships between annotations and foreground objects with trajectories. As shown in the experiments, the proposed approach can arrange annotations based on the moving trajectories of foreground objects and prevent the occlusions between different annotations and foreground objects. It also achieves better quantitative and qualitative results compared with state-of-the-art approaches.

Trajectories provide spatial-temporal information of foreground objects for event clustering and analysis. Because of the kinematic properties of foreground objects, the lengths of trajectories will be different which lead to the length problem of assessing similarity between two or more trajectories. To solve the problem, we propose a novel descriptor named trajectory kinematics descriptor to represent trajectories based on the kinematic properties from the point-of-view of Frenet-Serret frames. As shown in the experiments, applying the proposed trajectory kinematics descriptor for event clustering can achieve better F-measure scores compared to the state-of-the-art methods.

Smart cities deploy various sensors such as microphones and RGB cameras to collect data to improve the safety and comfort of the citizens. As data annotation is expensive, self-supervised methods such as contrastive learning are used to learn audio-visual representations for downstream tasks. Focusing on surveillance data, we investigate two common limitations of audio-visual contrastive learning: false negatives and the minimal sufficient information bottleneck. Irregular, yet frequently recurring events can lead to a considerable number of false-negative pairs and disrupt the model’s training. To tackle this challenge, we propose a novel method for generating contrastive pairs based on the distance between embeddings of different modalities, rather than relying solely on temporal cues. The semantically synchronized pairs can then be used to ease the minimal sufficient information bottleneck along with the new loss function for multiple positives. We experimentally validate our approach on real-world data and show how the learnt representations can be used for different downstream tasks, including audio-visual event localization, anomaly detection, and event search. Our approach reaches similar performance as state-of-the-art modality- and task-specific approaches.

Installing audio-based applications exposes users to the risk of the data processor extracting additional information beyond the task the user permitted. To solve these privacy concerns, we propose to integrate an on-edge data obfuscation between the audio sensor and the recognition algorithm. We introduce a novel privacy loss metric and use adversarial learning to train an obfuscator. Contrary to existing work, our technique does not require users to specify which sensitive attributes they want to protect (opt-out) but instead only provide permission for specific tasks (opt-in). Moreover, we do not require retraining of recognition algorithms, making the obfuscated data compatible with existing methods. We experimentally validate our approach on four voice datasets and show that we can protect several attributes of the speaker, including gender, identity, and emotional state with a minimal recognition accuracy degradation.