Research And System Application Of Face Recognition Based On Deep Learning

Face recognition technology is one of the hot topics in the field of computer vision,which is widely used in e-commerce,public security and smart cities.Face images in real scenes are prone to non-ideal light face images,side face images,and face images with exaggerated expressions,resulting in a decline in recognition accuracy.In recent years,deep learning has been widely used in the fields of image enhancement and noise reduction.In order to obtain the face image under ideal illumination from the non-ideal illumination face image,so as to improve the recognition accuracy,this paper proposes a deeper face image enhanced convolution network under non-ideal illumination through deep learning and develops a face recognition attendance system.The main contents of this article are as follows:(1)For face images under non-ideal lighting,a face enhancement convolutional neural network based on Retinex is proposed.The network structure is designed according to the Retinex theory.At the same time,in order to retain the characteristic information such as face contour edges,the direction gradient loss is combined with the Euclidean distance loss as the total network loss function.Experiments on the Yale dataset and the CMU-PIE dataset show that the algorithm can effectively obtain ideally illuminated face images,and through face recognition simulation experiments,the results show that the ideally illuminated face images generated by the network are effective Improves the accuracy of face recognition under non-ideal lighting.(2)Aiming at the monitoring environment with ordinary cameras,based on micro-service technology to integrate facial algorithm services and facial data management services,a set of face recognition attendance system was designed and developed.The system uses Golang,a highly concurrent language,and a face algorithm based on deep learning for development and deployment.The system deployment test shows that the system guarantees the reliability and effectiveness of the face recognition attendance method.(3)For the development and deployment of the system,Docker technology is adopted to ensure the consistency of the development and deployment environment,and it is easier to maintain;for the interactive mode of the application,ReactJS is used to develop a single multi-platform application program;for the system to input a large number of face images In the output,an image server based on the Docker engine was deployed using Nginx.This architecture design ensures the processing performance of the algorithm server.