Suspicious Activity and Theft Detection Using Deep Learning
P Suganthi, A Jaganaath, J Dhyanesh, A Aravindan · 2024
In order to improve their monitoring capabilities, smart CCTV systems depend on detecting anomalous events, human behaviour, and objects. Using these tools, we can keep tabs on our surroundings, spot suspicious humanbehaviour, and identify out-of-the-ordinary occurrences. Identifying and detecting anomalies in CCTV video streams is a common task for machine-vision and machine-learning algorithms. In most cases, these systems employ supervised learning to teach their algorithms to process video frames separately. The system is being trained using unsupervised and semi-supervised learning approaches due to the diversity and difficulty of anomalies. These methods reduce or eliminate the need for human intervention in the identification and creation of alarms in response to suspicious occurrences in CCTV feeds. Furthermore, by preserving routine scenarios at a lesser quality and abnormal events at their original quality, the system’s storage efficiency is enhanced.