Cognitive Video Analysis for Anomaly Detection Using Deep Learning

In academic research, deep learning behavior detection in films is essential for establishing a safe environment by automatically distinguishing between normal and questionable activity. To stop any financial and security risks, they take a proactive approach to spotting anomalous trends. To better understand how deep learning techniques might be applied to cognitive video analysis, this research article focuses on the identification of unusual human behavior in university surveillance footage. The work tackles the difficulties of detecting suspicious activity in the field of video content analytics (VCA), using deep learning to distinguish between typical and anomalous behavior. Based on learning measurements and objectives, the study offers a thorough analysis of deep learning-based methodologies. Three models-ConvLSTM2D, LRCN (Long-term Recurrent Convolutional Network and MobileNetV2 models are implanted for anomaly detection. The accuracy of ConvLSTM2D model is 86%, LRCN model is 90% and MobileNetV2 model is 92%.