Detection of Surgical Instruments Based on YOLOv5

With the development of science and technology, minimally invasive surgery has gradually played an important role in the medical field and has become the primary choice of all kinds of surgery. Compared with traditional surgery, minimally invasive surgery is simpler, less burden on doctors during surgery, and less pain, traumas and recovers rapidly after surgery. However, when having minimally invasive surgery, doctors cann't directly see inside of the body, and the actual operating space is small, which reduces doctors' coordination ability of hands and eyes. It may lead to the damage of surgical instruments or secondary injury to the internal tissues and organs of patients during surgery. Therefore, it needs reliable visual detection to monitor the process of surgery and improve the safety of surgery. In this paper, we propose a real-time detection model of surgical instruments based on YOLOv5. We selected a real and public dataset for training and verifying, and through experiments, we calculated precision, recall and mAP to evaluate the performance of the model.