ENHANCED HUMAN ACTIVITY RECOGNITION FOR SECURITY SURVEILLANCE USING DEEP NEURAL NETWORKS

Motion detection in live video streams is considerably simpler when deep learning is used. Security monitoring is improved as a result. In this research, we investigate how different spatial and temporal features are analyzed by Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), including Long Short-Term Memory (LSTM) models, to look for anomalies in surveillance footage. Deep learning-based activity recognition frequently performs better than more traditional machine learning techniques. As a result, security systems become more reliable and responsive. The ability of this approach to make correct behavior judgments could revolutionize automated tracking and public safety by enabling more proactive and efficient surveillance.