AI-Based Real-Time Violence Detection System For Social Media Content Moderation

Abstract

The social media has evolved at an exceptionally high rate, which has promoted the proliferation of violent and abusive messages that present a significant threat to internet safety and the social welfare of the community. Manual moderation is ineffective, subjective and cannot deal with the huge number of uploads that are made every day. The proposed project is a real-time violence detection system, which is an AI-based application that should automatically recognize harmful material in video, audio, and text. The system is a combination of computer vision, natural language processing, and deep learning to make the accuracy high. Video data are analyzed by extracting frames and classifying them, analyzing audio by spectrogram analysis, determining the presence of aggressive sounds, and analyzing the text to determine the presence of hate speech or violence. Visualization is used to enhance the level of model interpretability, and real-time moderation is provided by a web-based interface. Findings indicate more efficient and secure Internet spaces.