GDLAVID-graph-based Deep Learning Approach for Automatic Violence Detection in Videos

This paper presents a method for detecting violence in videos using Graph Neural Networks (GNNs) and Spatio-Temporal Graph Neural ‎Networks (ST-GNNs). In this approach, each video frame is turned into a graph where people and objects are treated as nodes, and their ‎interactions are represented by connections. By studying these interactions over time, violent activities can be identified. The method was ‎tested on the Smart-City CCTV Violence Detection Dataset for Automatic Violence Detection in Videos, from Kaggle, which contains short ‎video clips labeled as violent or non-violent. The results show that this technique is effective in recognizing violent incidents in different ‎situations, making it useful for public safety and real-time surveillance.