Abstract:Video anomaly detection is generally known as automatic recognition of video frames that contain abnormal behaviors or events. It has been broadly utilized in many applications, such as surveillance and public safety. However, video anomaly detection remains a challenging issue due to the complexity of real data and the difficulty of defining anomaly unambiguously. In current prediction methods, the use of motion characteristics is insufficiency, and the structural information is not fully retained. In this paper, we thus propose an innovative framework for video anomaly detection based on a dynamic convolutional generative adversarial network. Our framework is composed of a frame prediction network and a convolution-based discrimination network in a GAN-like architecture. The prediction network equipped with a convolutional dynamic neural advection unit, which can maximize the retention of structural and motion information in the predicted frames to obtain high-quality future frames. At the same time, we exploit a 3D convolutional approach to assemble the discrimination network, which can not only receive the whole video sequence at once, but also extract the spatio-temporal features simultaneously. Additionally, we leverage a novel combination loss to enforce the consistency of video frame appearance and motion. Under the weakly supervised paradigm, our model trains exclusively on normal samples. Experimental results on three benchmark datasets show that our approach achieves good effectiveness and robustness compared with state-of-the-art methods. Remarkably, our model obtains the AUC score of 95.3% and 75.2% on the Ped2 and ShanghaiTech dataset.

Non-local Attention Based Generative Adversarial Network for Video Abnormal Event Detection

Generative Adversarial Networks for Abnormal Event Detection in Videos Based on Self-Attention Mechanism

Video Anomaly Detection Based on Frame Prediction of Generative Adversarial Network

Self-Supervised Attentive Generative Adversarial Networks for Video Anomaly Detection

Abnormal Event Detection Algorithm Based on Dual Attention Future Frame Prediction and Gap Fusion Discrimination.

A Dynamic Convolutional Generative Adversarial Network for Video Anomaly Detection

Surveillance Video Anomaly Detection Via Non-Local U-Net Frame Prediction

3D U-Net for Video Anomaly Detection.

Learning Normal Patterns Via Adversarial Attention-Based Autoencoder for Abnormal Event Detection in Videos

Improved Video Anomaly Detection with Dual Generators and Channel Attention

Tam-Net: Temporal Enhanced Appearance-To-Motion Generative Network For Video Anomaly Detection

Video Abnormal Event Detection Via Context Cueing Generative Adversarial Network

Video Anomaly Detection Using Dual Discriminator Based Generative Adversarial Network.

Detecting Anomalies in Videos Using Perception Generative Adversarial Network.

Anomaly Detection in Traffic Surveillance Videos with GAN-based Future Frame Prediction

Detecting abnormality with separated foreground and background: Mutual Generative Adversarial Networks for video abnormal event detection

Convolutional Transformer based Dual Discriminator Generative Adversarial Networks for Video Anomaly Detection

Attention-guided generator with dual discriminator GAN for real-time video anomaly detection

Learning Attention Augmented Spatial-temporal Normality for Video Anomaly Detection

Multi-Objective Optimized Generative Adversarial Networks for Video Anomaly Detection

Context-related video anomaly detection via generative adversarial network