An Enhanced Surveillance System for Gun and Knife Detection using YOLOv8 and Raspberry Pi

In response to growing concerns surrounding public safety, this research proposes an innovative surveillance system designed to detect firearms and knives in real-time using a customized YOLOv8 (You Only Look Once version 8) object detection algorithm. The system is implemented on a Raspberry Pi platform with a Picam for efficient and cost-effective deployment in diverse environments. The primary motivation is to address the limitations of existing surveillance systems, offering an enhanced solution capable of accurately identifying potential threats and alerting relevant authorities promptly. Recognizing the need for improved systems, we delve into the methodology, starting with the collection of a diverse dataset crucial for training the YOLOv8 model. The dataset encompasses various scenarios, lighting conditions, and angles to ensure the robustness of the model. The YOLOv8 algorithm is then tailored to detect guns and knives, involving a meticulous training process. The real-time object detection capability of the system is thoroughly examined, encompassing the capture of video frames and the execution of the YOLOv8 model on the Raspberry Pi. Challenges encountered during implementation are discussed, accompanied by effective solutions to ensure a reliable and responsive surveillance system. An integral component of the proposed system is the alerting mechanism, which promptly notifies relevant authorities or users upon the detection of firearms or knives. To evaluate the system’s performance, extensive testing is conducted under diverse scenarios, assessing its accuracy, precision, and recall metrics.