Contracting Skeletal Kinematics for Human-Related Video Anomaly Detection
This work addresses the problem of detecting rare anomalous human behaviors, such as fights or falls, for applications like surveillance and safety, though it is incremental in improving existing methods with novel latent spaces.
The paper tackles human behavior anomaly detection in videos by proposing COSKAD, a model that encodes skeletal motion with a graph convolutional network and contracts embeddings onto a minimal-volume latent hypersphere, achieving state-of-the-art results on datasets like UBnormal, ShanghaiTech Campus, and CUHK Avenue.
Detecting the anomaly of human behavior is paramount to timely recognizing endangering situations, such as street fights or elderly falls. However, anomaly detection is complex since anomalous events are rare and because it is an open set recognition task, i.e., what is anomalous at inference has not been observed at training. We propose COSKAD, a novel model that encodes skeletal human motion by a graph convolutional network and learns to COntract SKeletal kinematic embeddings onto a latent hypersphere of minimum volume for Video Anomaly Detection. We propose three latent spaces: the commonly-adopted Euclidean and the novel spherical and hyperbolic. All variants outperform the state-of-the-art on the most recent UBnormal dataset, for which we contribute a human-related version with annotated skeletons. COSKAD sets a new state-of-the-art on the human-related versions of ShanghaiTech Campus and CUHK Avenue, with performance comparable to video-based methods. Source code and dataset will be released upon acceptance.