Face Recognition with Siamese Networks

Abstract Face recognition method is a common problem in modern Internet life. In this paper, face recognition will be designed based on Siamese network. The traditional methods have some problems, such as being greatly affected by environmental factors, insufficient processing of individual differences, lack of robustness and adaptability. However, the existing Siamese network based method has better recognition effect. In this paper, a face recognition technology based on Siamese network will be proposed. In this method, the features of two face images are extracted by the same neural network, and then these feature vectors are input into a distance metric model to calculate the similarity or distance between them, which is used to compare the similarity between faces. This method reduces the training parameters of the model, improves the utilization of parameters, and the average accuracy can reach 98%. Contrasted with traditional face recognition techniques, it offers enhanced accuracy and resilience, delivering favourable outcomes in real-world applications.