Introduction - If you have any usage issues, please Google them yourself
The algorithm model of facenet face recognition is obtained through deep learning, and the backbone network of feature extraction is concept-resnetv1, which is developed from concept network and RESNET, with more channels and network layers, so that each layer can learn more features and greatly improve the generalization ability. The network is deeper, the amount of calculation in each layer is reduced, and the ability of feature extraction is strengthened, so as to improve the accuracy of target classification. On the LFW data set, the accuracy of face recognition reaches 98.40%. In this experiment, mtcnn is introduced into the face detection algorithm. Its backbone network is divided into three convolutional neural networks: p-net, R-Net and o-net. Among them, o-net is the most strict in screening candidate face frames. It will output the coordinates of a human face detection frame and five facial feature points (left eye, right eye, nose, left mouth corner, right mouth corner).