TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Anomaly Detection	ShanghaiTech	STPT	AUC	77.1%	# 15
Anomaly Detection	ShanghaiTech	STPT	RBDC	51.6	# 2
Anomaly Detection	ShanghaiTech	STPT	TBDC	84.6	# 5
Anomaly Detection	UCSD Ped2	STPT	AUC	98.9%	# 2

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spatio-temporal-predictive-tasks-for-abnormal/anomaly-detection-on-ucsd-ped2)](https://paperswithcode.com/sota/anomaly-detection-on-ucsd-ped2?p=spatio-temporal-predictive-tasks-for-abnormal)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/spatio-temporal-predictive-tasks-for-abnormal/anomaly-detection-on-shanghaitech)](https://paperswithcode.com/sota/anomaly-detection-on-shanghaitech?p=spatio-temporal-predictive-tasks-for-abnormal)`

Spatio-temporal predictive tasks for abnormal event detection in videos

27 Oct 2022 · Yassine Naji, Aleksandr Setkov, Angélique Loesch, Michèle Gouiffès, Romaric Audigier ·

Abnormal event detection in videos is a challenging problem, partly due to the multiplicity of abnormal patterns and the lack of their corresponding annotations. In this paper, we propose new constrained pretext tasks to learn object level normality patterns. Our approach consists in learning a mapping between down-scaled visual queries and their corresponding normal appearance and motion characteristics at the original resolution. The proposed tasks are more challenging than reconstruction and future frame prediction tasks which are widely used in the literature, since our model learns to jointly predict spatial and temporal features rather than reconstructing them. We believe that more constrained pretext tasks induce a better learning of normality patterns. Experiments on several benchmark datasets demonstrate the effectiveness of our approach to localize and track anomalies as it outperforms or reaches the current state-of-the-art on spatio-temporal evaluation metrics.

PDF Abstract