Real-Time Graffiti Vandalism Detection: Enhancing Public Safety with the YOLOv8 Algorithm

Graffiti vandalism, often considered a form of public expression, poses significant challenges to urban aesthetics and public safety. This study addresses the urgent need for effective real-time detection of graffiti vandalism using the YOLOv8 algorithm. The model was developed and evaluated across three distinct parameter scenarios, yielding a comprehensive comparison for identifying the optimal configuration for graffiti detection. Notably, Scenario 2 demonstrated the best performance, achieving a precision of 68,4% and recall of 65,4%, with a mean Average Precision (mAP) of 0.710. These quantitative metrics illustrate the effectiveness of the model for accurately detecting vandalism in various urban environments. By integrating a diverse dataset sourced from web scraping and existing open-source repositories, this study ensures robustness in detecting different graffiti types under various conditions. The findings indicate that the YOLOv8 model not only enhances the capability of authorities to monitor and manage graffiti effectively but also contributes to urban management systems by improving public safety and maintaining the aesthetic quality of public spaces.