The Transfer Learning Models for Face Recognition: A Survey

Facial recognition has many uses in marketing, biometric security, and healthcare. Facial recognition has emerged as the most attractive field of study among other biometrics, such as fingerprints, iris, and voice. Increasing safety and security, avoiding crime, minimizing human interaction, and even supporting medical efforts are just a few of the advantages that facial recognition technology provides for society. Several face recognition methods have been introduced, including Fisher faces, eigenvectors, and the local binary pattern graph (LBPH). Because of the low accuracy of these methods, a convolutional neural network is used because it has been proven through many experiments that it is the best among other algorithms in terms of accuracy and time. Computer-based face recognition is an established and trustworthy technology widely used in many access control settings. Transfer learning is critical for addressing the underlying problem of insufficient training data in machine learning. It moves knowledge from the source domain to the destination domain. This paper compares previous work using learning models with deep learning algorithms. This paper aims to explain why transfer learning is used to identify faces and to describe the characteristics and advantages of these transfer learning models, in addition to comparing the previous studies on face recognition and the algorithms used in devising the best methods to obtain high-accuracy results when recognizing the face.