Deep Learning-Based High-Accurate Violence Detection for Intelligent Video Surveillance Systems
Sumeet Kothari · Panamerican mathematical journal. · 2025
The rapid increase in violent incidents necessitates the development of intelligent video surveillance systems capable of real-time violence detection. This research introduces a high-accuracy deep learning framework integrating CNN and Bi-LSTM, significantly enhancing the detection of violent activities in surveillance footage. The proposed model outperforms existing approaches, achieving 99% accuracy, surpassing traditional deep learning models such as CNN, VGG16, and ResNet50. Extensive evaluations validate its robustness, precision, and recall, ensuring reliable monitoring with minimal false positives. The study highlights the role of deep learning in revolutionizing crime prevention by enabling automated, real-time security solutions.