Date of Publication :21st May 2024

Abstract:Automated violence detection in real-life scenarios is a pressing concern with profound implications for public safety, law enforcement, and societal well-being. This paper presents a comprehensive approach to violence detection utilizing deep learning techniques, particularly Convolutional Recurrent Neural Networks (CRNN), applied to video data. We commence with a thorough literature review, encompassing traditional and modern methods for violence detection, and underscore the limitations of existing approaches. Our methodology entails feature extraction from video frames and the utilization of a CRNN architecture, which amalgamates convolutional and recurrent neural network layers, leveraging the VGG19 model for feature extraction. We delineate the dataset employed for training and evaluation, underscoring the significance of data diversity and preprocessing techniques tailored to sequential data. Through meticulous experimentation, we demonstrate the effectiveness of our CRNN-based approach in accurately discerning instances of violence in videos. Our findings unveil promising performance metrics, including accuracy, precision, recall, and F1-score, underscoring the viability of our system for real-world deployment. Furthermore, we deliberate on the ethical ramifications of automated violence detection systems and outline future research trajectories, emphasizing the pivotal role of such systems in addressing contemporary societal challenges. In sum, this paper advances the state-of-the-art in violence detection and furnishes valuable insights for researchers and practitioners in the domains of computer vision and artificial intelligence.

Reference :