Advancements and Challenges of Deep Learning in Facial Recognition

Abstract. Facial recognition technology allows users to achieve efficient and accurate identity verification through facial features alone, without the need to physically touch the device. Face recognition technology can be applied in a variety of scenarios such as unlocking mobile devices, online payment, attendance management and public identity verification. Nonetheless, face recognition algorithms still have some limitations in the face of massive data and complex application scenarios. Therefore, there is an urgent need for more advanced technical means to overcome the shortcomings of traditional algorithms such as low recognition rate and poor adaptability. As deep learning methods have become more popular, especially the emergence of Convolutional Neural Networks (CNN) and Transformer Networks, face recognition has been revolutionized. This paper firstly describes the principles of convolutional neural network model and Transformer model. Secondly, the applications of CNNs and Transformer networks at different stages of development in the domain of face recognition are reviewed. Then, this paper combs through the deep learning models applied to face recognition methods and evaluates their characteristics, innovativeness, usefulness and portability respectively. Finally, this paper summarizes the challenges faced by deep learning in the domain of face recognition as well as the trend of face recognition technology development in the future, aiming to point out the research focus research direction for the subsequent research.