One-Shot Learning for Face Recognition, Method Comparison

Abstract

Deep learning uses multiple layers to represent data at different levels to extract the required information. Where this technology flourished and became an important area of research direction in 2014 after its use to prevent the penetrations of Deep-ID and Deep-face. This technology is characterized by a hierarchical shape that works on merging the pixels together to form the image of the face. This has led to a significant improvement in advanced performance and an effective contribution to the success of applications in the world. As for facial recognition in one case, it is a unique case that characterizes the human race, and researchers have tried to create one-shot deep learning algorithms for the purpose of simulating this ability. Despite the excellent performance of deep learning and its methods for classifying different image problems, its performance often depends on the presence of a large number of training samples. In this research, we compare a face-recognition methodology utilizing neural networks CNN but using different databases through published research in this field, and the focus was also on one-shot learning because it is one of the great challenges at the present time in order to be a simplified reference for researchers in the field of one-shot face recognition and the key to launching their research.