Video-based Anomalous Activity Detection Using 3D-CNN and Transfer Learning

Video surveillance is used to observe various behaviour. It is used for security purpose. CCTV cameras are used for detecting the anomaly. The transfer learning is used to detect any unusual behaviour by human. The collected by the CCTV can also be used for detecting anomalous activity from the recorded video. By extracting the features of a target object’s movement and status, and the use of transfer learning helps to recognize activities or unusual action by the human and describe them in natural language.Such a surveillance system can detect, track, and analyse any unusual behaviour of the person in high security needed areas or from the crowded scenes by using video data. Anomaly detection by the visual appearance and motion dynamics of the involved humans and objects in video sequences is explained here with the help of transfer learning.