Application of Face Data Augmentation Based on Rotate-and-Render-DCGAN in Campus Security

Targeting at solving the problem of low accuracy of existing recognition methods caused by the imbalance of face data sets in the university security system, an unbalanced Data Augmentation method based on Rotate-and-Render-DCGAN is proposed to improve face recognition accuracy campus. Our goal is to correct the angled faces in the data set through Rotate-and-Render and generate artificial images to enrich the image database combined with DCGAN to improve the classifier's performance and the accuracy of face recognition. We evaluate and compare the impact of traditional Data Augmentation and Rotate-and-Render- DCGAN Data Augmentation on face recognition accuracy. The results show that the improvement effect of this method is more significant.

I. Introduction

Colleges and universities have a unique regional model, the surrounding environment is involved, the school's safety awareness is reduced, the mobility of personnel in the school is significant, the safety management personnel are few, and the inspection scope is broad. [1], [2] Therefore, that is of considerable significance to strengthen campus safety management, take practical and effective measures to protect the safety and rights and interests of college students, and ensure the all-round development of their physical and mental health. Therefore, it is urgent to improve the intelligent campus security system.[3], [4] Improving the unmoving face's automatic recognition function in the smart campus security system is essential in meeting the smart campus security system.