Published Paper Details:

With the exponential growth of video surveillance systems in public and private spaces, automatic video anomaly detection has become a critical research area. Manual monitoring of surveillance footage is labor-intensive and prone to human error. Deep learning, particularly neural networks, has revolutionized video anomaly detection through robust spatiotemporal feature learning. This survey presents a comprehensive analysis of recent advancements in neural network-based methods for anomaly detection in video surveillance. We discuss foundational approaches, evaluate cutting-edge techniques such as autoencoders, GANs, transformers, and skeletal-based models, and highlight key datasets and performance metrics. Comparative analyses demonstrate the trade-offs across architectures and detection paradigms. The paper concludes with future research directions in privacy-preserving models, multimodal fusion, and explainable AI for real-world deployments.