Abstract
Facial recognition research has achieved significant progress in recent years. State-of-the-art facial recognition models have already passed the level of human accuracy. In the meantime, facial recognition is already used in many fields, including making purchases, authenticating mobile applications, verifying identities in online shopping, and automating border control in airports. This paper compares different database options based on different tasks and use cases of facial recognition applications from the perspective of scalability. In particular, a novel approach to graph databases is proposed to detect false positive misclassifications in facial recognition. Furthermore, the latest homomorphic encryption schemes enable cloud-based facial recognition to run with encrypted data without requiring private keys. In addition to providing a guide for researchers and studies on facial recognition systems, the purpose of this article is also to create a perspective for the use of the most appropriate system.
