AUTOMATICALLY DETECT SUSPICIOUS HUMAN ACTIVITY BY USING VARIOUS TOOLS AND TECHNOLOGIES (KEY IJP************692)

• Raj Kumar Gupta

Abstract

Detecting suspicious behavior using a combination of deep learning techniques, particularly LRCN (Long-term Recurrent Convolutional Network). This method allows for the analysis of temporal data in video frames, which is crucial for identifying anomalies in human activity.Using a combination of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), the model can effectively extract relevant features from video data and classify behavior as suspicious or not. The process involves several key stages, including research, data collection and preprocessing, model design and training, and performance evaluation.It's great that we've utilized datasets like KTH and Kaggle to train and validate your model. By leveraging these resources, you've been able to achieve an impressive accuracy of 86% in detecting suspicious events. Additionally, as you continue to expand your dataset, it's reasonable to expect further improvements in accuracy. Seems well-structured and promising for enhancing public safety and security through the automated detection of suspicious behavior in real-time video footage.