Comparison of Deep Learning Techniques for Violent Content Detection in Video Stream

Abstract

The widespread adoption of surveillance systems has underscored the need for automated violence detection methods to bolster public safety. This survey explores both conventional machine learning techniques and modern deep learning approaches employed for violence detection. It examines the progression of these methodologies, evaluates their effectiveness on standard datasets, and sheds light on current challenges, including the complexities of real-time analysis and ethical dilemmas. Additionally, the paper suggests prospective research avenues aimed at overcoming these obstacles and enhancing the accuracy and reliability of violence detection systems.