Leveraging Attention Mechanism to Enhance Culprit Identification in Real-Time Video Surveillance Using Deep Learning

This research presents an efficient approach to improve real-time video surveillance by integrating Attention Mechanisms (AM) within the deep learning model. The study aims to enhance the identification of culprits in surveillance footage. The attention mechanisms enable the model to focus on relevant regions of interest within the video frames, reducing computational overhead and increasing accuracy when compared with the existing works. The proposed work achieves 92% accuracy with an improvement of 0.07, 91% precision with an improvement of 0.03, 94% recall with an improvement of 0.12, and, 92% F1-Score with an improvement of 0.07, and contributes to the field of security and surveillance by providing an efficient and effective method for identifying suspicious activities or individuals, thereby increasing the capabilities of real-time video surveillance systems for enhanced security and safety applications.