Design of Face Recognition Fatigue Driving Detection System Based on Improved YOLO Algorithm

Face recognition is a very important step in driver fatigue driving detection, which is directly related to the subsequent driver's eye detection and location, the judgment of driver's eye state, etc. In recent years, DL (Deep learning) has made great progress in face recognition. For face recognition in natural scenes, researchers have made long-term research and substantial progress. In this paper, the research of face recognition fatigue driving detection system based on improved YOLO algorithm is carried out, and a face recognition fatigue driving detection system is established. In this paper, the YOLOv5 network structure is adopted, and the Label Smoothing method is introduced into the Prediction layer of YOLOv5. By defining a central loss function, the intra class variation is minimized while keeping different features separable, thus improving the inter class separability. The results show that the accuracy is 97.421% when the threshold is set to 0.5. The accuracy obtained is 4.871% higher than the original YOLOv5 algorithm. It can meet the requirements of practical application.