Abnormal Event Recognition is Surveillance using Multibranch- Generative Adversial Network

The detection of abnormal events in surveillance footage is crucial for public safety and security in various environments. Traditional methods, which often rely on human monitoring, are time-consuming and prone to error. Real- time anomaly detection can now be automated through deep learning technologies. Automated anomaly identification is made possible by Generative Adversarial Networks (GANs), an effective method for video analysis. This project introduces a novel approach using a Multi- Branch GAN (M-GAN) model specifically designed for detecting anomalies in surveillance video. The M-GAN model employs a two-stage process, learning the distribution of normal events and Identifying deviations as potential anomalies. Its key advantage is its ability to operate without labelled anomaly data, adapting more flexibly to various environments and conditions. Experimental results show that the M-GAN outperforms conventional GAN- based methods, achieving higher precision and recall rates in detecting abnormal events. This robust performance positions M-GAN as a leading solution for real- time abnormal event detection in surveillance systems, promising to enhance safety and security while reducing the need for extensive manual oversight.