Detection of Violent Content in Videos using Audio Visual Features
Rishab KS, P Mayuravarsha, Yashwal S Kanchan, Pranav MR, Roopa Ravish · 2023
Content moderation is an essential part of the internet, which makes the internet a safe place for people of all age groups. Content moderation is usually done manually by humans, who flag the content as violent or non-violent. The content moderators are exposed to various violent content, which affects them mentally and emotionally. In this research, we introduce a machine learning model that can identify and categorizing videos as either violent or non-violent based on the presence of violent content. Two parameters are taken as inputs namely audio and video. The audio is separated from the video initially. The audio component is processed through an audio classifier which distinguishes between violent and nonviolent sounds. If the audio classifier classifies audio as nonviolent, then the corresponding video is fed into a video classifier where it is categorized into violent or non-violent video.