TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Video Anomaly Detection	HR-Avenue	MPED-RNN	AUC	86.3	# 7
Video Anomaly Detection	HR-ShanghaiTech	MPED-RNN	AUC	75.4	# 6
Video Anomaly Detection	HR-UBnormal	MPED-RNN	AUC	61.2	# 6
Anomaly Detection	ShanghaiTech	MPED-RNN	AUC	73.40%	# 20

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-regularity-in-skeleton-trajectories/video-anomaly-detection-on-hr-shanghaitech)](https://paperswithcode.com/sota/video-anomaly-detection-on-hr-shanghaitech?p=learning-regularity-in-skeleton-trajectories)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-regularity-in-skeleton-trajectories/video-anomaly-detection-on-hr-ubnormal)](https://paperswithcode.com/sota/video-anomaly-detection-on-hr-ubnormal?p=learning-regularity-in-skeleton-trajectories)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-regularity-in-skeleton-trajectories/video-anomaly-detection-on-hr-avenue)](https://paperswithcode.com/sota/video-anomaly-detection-on-hr-avenue?p=learning-regularity-in-skeleton-trajectories)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-regularity-in-skeleton-trajectories/anomaly-detection-on-shanghaitech)](https://paperswithcode.com/sota/anomaly-detection-on-shanghaitech?p=learning-regularity-in-skeleton-trajectories)`

Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos

CVPR 2019 · Romero Morais, Vuong Le, Truyen Tran, Budhaditya Saha, Moussa Mansour, Svetha Venkatesh ·

Appearance features have been widely used in video anomaly detection even though they contain complex entangled factors. We propose a new method to model the normal patterns of human movements in surveillance video for anomaly detection using dynamic skeleton features. We decompose the skeletal movements into two sub-components: global body movement and local body posture. We model the dynamics and interaction of the coupled features in our novel Message-Passing Encoder-Decoder Recurrent Network. We observed that the decoupled features collaboratively interact in our spatio-temporal model to accurately identify human-related irregular events from surveillance video sequences. Compared to traditional appearance-based models, our method achieves superior outlier detection performance. Our model also offers "open-box" examination and decision explanation made possible by the semantically understandable features and a network architecture supporting interpretability.