An Attribute-based Method for Video Anomaly Detection

About

Video anomaly detection (VAD) identifies suspicious events in videos, which is critical for crime prevention and homeland security. In this paper, we propose a simple but highly effective VAD method that relies on attribute-based representations. The base version of our method represents every object by its velocity and pose, and computes anomaly scores by density estimation. Surprisingly, this simple representation is sufficient to achieve state-of-the-art performance in ShanghaiTech, the most commonly used VAD dataset. Combining our attribute-based representations with an off-the-shelf, pretrained deep representation yields state-of-the-art performance with a $99.1\%, 93.7\%$, and $85.9\%$ AUROC on Ped2, Avenue, and ShanghaiTech, respectively.

Tal Reiss, Yedid Hoshen · 2022

Related benchmarks

Task	Dataset	Result
Video Anomaly Detection	UCF-Crime	AUC—	288
Video Anomaly Detection	ShanghaiTech (test)	AUC—	211
Video Anomaly Detection	UCF-Crime (test)	AUC—	189
Abnormal Event Detection	UCSD Ped2	AUC—	163
Abnormal Event Detection	UCSD Ped2 (test)	AUC—	146
Video Anomaly Detection	Avenue (test)	AUC (Micro)93.7	85
Video Anomaly Detection	CUHK Avenue	Frame AUC93.6	75
Video Anomaly Detection	ShanghaiTech	ROC AUC—	61
Video Anomaly Detection	UBnormal	AUC—	60
Video Anomaly Detection	UBnormal (test)	AUC—	55