Video anomaly detection is a real-world problem that artificial intelligence (AI) and computer vision applications can solve. In today's world, with our deteriorating environment, it is urgent to assess surveillance videos and process them in real-time for abnormalities to ensure the safety and security of citizens. Several deep-learning approaches have been developed to detect anomalies in videos. However, these traditional models require improvements in hybridization and utilize a Generative Adversarial Network (GAN) architecture to achieve enhanced performance. This paper presents a novel deep-learning framework that efficiently addresses this problem by searching surveillance footage for irregularities. A deep learning technique called GANDL-VAD is suggested for Video Abnormal Detection (VAD). It uses a GAN architecture and offers a hybrid DL model that considers both extracted and synthesized data, thereby improving the efficiency of detection and classification. The suggested hybrid deep learning model surpassed its modern rivals with an accuracy rate of 98.78%, according to experimental results on the UCF-Crime benchmark dataset. It can be used directly in current computer vision applications for video analytics and is capable of detecting anomaly events in surveillance videos.
A Thu, study studied this question.