Enhanced home security system: Anomaly and weapon detection, surveillance, and police coordination using SVM, random forest, YOLO
Venkata Praneeth Gaddam, Aasam Sri Ram, Devi M. Anousouya · 2025
Since video surveillance systems are being used more often across a wider range of industries, it is imperative to accurately detect human anomalies in order to maintain public safety. Unsupervised deep learning models have become a viable method for automatically detecting anomalous human behavior in video surveillance under this particular set of circumstances. The several unsupervised deep learning methods that have been developed for anomaly identification in video surveillance are thoroughly reviewed in this study. This paper covers a variety of models, such as variational autoencoders (VAEs), generative adversarial networks (GANs), and autoencoders. Without labeled training data, these models can learn the innate representation of typical behavior patterns in movies and identify departures from these patterns. The study also discusses each model’s main benefits and drawbacks and offers information on how well each model works in various video surveillance scenarios. The difficulties and potential paths forward for enhancing the efficacy and efficiency of unsupervised deep learning models for anomaly detection in video surveillance are also covered in the review. All things considered, this evaluation provides a thorough resource for scholars and professionals who are trying to create effective and trustworthy techniques for identifying anomalies in human behavior in video surveillance systems.