Research on FaceNet Face Recognition Algorithm Based on Attention Mechanism

FaceNet face recognition algorithm is the mainstream face recognition algorithm at present, and its running speed is widely used in the industry. To further improve the accuracy of FaceNet face recognition network, a FaceNet face recognition algorithm based on attention mechanism is proposed. Based on FaceNet, the algorithm introduces a Convolutional Block Attention Module (CBAM) attention mechanism to strengthen the feature extraction network to amplify the local information in the feature maps of the three scales, strengthen the feature extraction under different receptive fields, and enhance the more important feature information. Cross-Entropy Loss and Triplet Loss are combined as joint loss functions, and a classifier is used to assist the training so that the training is more convergent. The experimental results show that the proposed face recognition algorithm achieves a good recognition effect, with an accuracy of 99.50% on the face data set (LFW).