Video anomaly detection and localization via Gaussian Mixture Fully Convolutional Variational Autoencoder

We present a novel end-to-end partially supervised deep learning approach for video anomaly detection and localization using only normal samples. The insight that motivates this study is that the normal samples can be associated with at least one Gaussian component of a Gaussian Mixture Model (GMM),...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Computer vision and image understanding Ročník 195; s. 102920
Hlavní autori: Fan, Yaxiang, Wen, Gongjian, Li, Deren, Qiu, Shaohua, Levine, Martin D., Xiao, Fei
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Elsevier Inc 01.06.2020
Predmet:
Anomaly detection
Dynamic flow
Gaussian mixture model
Two-stream network
Variational autoencoder
Video surveillance
Variational autoencoder
Video surveillance
Dynamic flow
Two-stream network
Anomaly detection
Gaussian mixture model
ISSN:1077-3142, 1090-235X
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:We present a novel end-to-end partially supervised deep learning approach for video anomaly detection and localization using only normal samples. The insight that motivates this study is that the normal samples can be associated with at least one Gaussian component of a Gaussian Mixture Model (GMM), while anomalies either do not belong to any Gaussian component. The method is based on Gaussian Mixture Variational Autoencoder, which can learn feature representations of the normal samples as a Gaussian Mixture Model trained using deep learning. A Fully Convolutional Network (FCN) that does not contain a fully-connected layer is employed for the encoder–decoder structure to preserve relative spatial coordinates between the input image and the output feature map. Based on the joint probabilities of each of the Gaussian mixture components, we introduce a sample energy based method to score the anomaly of image test patches. A two-stream network framework is employed to combine the appearance and motion anomalies, using RGB frames for the former and dynamic flow images, for the latter. We test our approach on two popular benchmarks (UCSD Dataset and Avenue Dataset). The experimental results verify the superiority of our method compared to the state of the art. •An end-to-end VAE framework has been considered for video anomaly detection.•A two-stream network employs dynamic flows for detecting the motion anomalies.•Sample energy method is applied to detect anomalies with probabilities of the GMM.•Experiment result on two public datasets demonstrate the superiority of our method.
ISSN:1077-3142
1090-235X
DOI:10.1016/j.cviu.2020.102920