Face recognition in images based on deep learning

In view of the mutual interference between student distribution, perspective, scale and target in the classroom environment and the uncertainty of lighting, face detection in the classroom is faced with a large error. Through in-depth study of YOLOV5s architecture, CBAM attention mechanism was embedded into SPPF module of the pattern skeleton, and C3 module was transformed into C2. The accuracy rate of this model was $70.1 \%$, the recall rate was $66.3 \%$, and the average accuracy was $66.9 \%$, which was $4 \%$ higher than the average accuracy of the original model.