Detection of Violent Elements in Digital Games Using YOLO V7 Algorithm

Abstract

Digital games have various useful purposes such as entertainment, education, attention strengthening, and strategy development, but they can also negatively affect users with harmful content such as violence, gambling, sexuality, profanity, and drugs. Especially for children at a young age, these elements can settle into their subconscious and negatively affect their lives in the future. Violence elements such as gun, knife, fights, fire, and blood are very common in digital games in many different forms. Real-time detection of these violence elements is crucial to enable early intervention in possible problems. In this study, violence elements within digital games were examined, and a deep learning method was proposed to detect these elements such as guns, knives, bombs, fire, blood, and fights. Additionally, the detection of these violence elements in digital games was performed using YOLOV7. The proposed model was trained on Google Colab, and test results showed that it could be used for real-time digital games with high accuracy performance and detection speed.