Aiming at the problems of missed detection and false detection of face detection in the current face detection model in complex environments, YOLOV5s-Face face detection algorithm is proposed. It is improved based on YOLOV5s. First, the anchor box's size is modified using the K-means algorithm, which determine the priority anchor box size, and improve the fitness of the anchor box and the actual object. Secondly, the SE attention mechanism is embedded in the backbone network structure of YOLOV5s. It might improve the network's capacity for feature extraction. Finally, this research employs four-scale feature detection to enhance the network's detection performance on small face targets. The outcomes of the experiments demonstrate that the YOLOV5s-Face algorithm has an increase of 3.9 percentage points and 4 percentage points in the mAP and Recall of the YOLOV5s algorithm respectively. YOLOV5s-Face algorithm can better detect face.