Violence and Robbery Detection System Using YOLOv5 Algorithm Based on IoT Technology

Violence and robbery are two common forms of crime that often cause material losses, psychological trauma, and insecurity within society. Conventional CCTV systems are limited in preventing such incidents, which highlights the need for more intelligent and responsive security solutions. The primary objective of this research is to design and evaluate SmartGuard, a real-time detection system for violence and robbery based on artificial intelligence (AI) using the YOLOv5 algorithm, integrated with Internet of Things (IoT) technology for remote monitoring. This study employed an experimental design with several stages: dataset preparation, model training, testing, model analysis, and system integration with Raspberry Pi, Firebase, and a mobile application. The dataset consisted of 6,900 labeled images across three classes: violence, robbery, and normal activity. Model evaluation was conducted using a separate test dataset and analyzed with a confusion matrix. The results show that the model achieved an overall accuracy of 70.94%. The system performed relatively well in detecting violence, with a precision of 71.13% and an F1-score of 62.47%. However, recall values for robbery (47.53%) and normal activity (48.99%) were considerably lower, indicating challenges in consistently recognizing these classes. Despite these limitations, SmartGuard allows users to view and receive notifications in emergency situations, enabling them to take quick action and monitor the situation effectively.