Deep learning based single sample face recognition: a survey

Face recognition has long been an active research area in the field of artificial intelligence, particularly since rise deep learning recent years. In some practical situations, each identity only a single sample available for training. under this situation is referred to as face and poses significant challenges effective training models. Therefore, years, researchers have attempted unleash more potential improve model performance situation. While several comprehensive surveys conducted on traditional approaches, emerging based methods are rarely involved these reviews. Accordingly, we focus learning-based paper, classifying them into virtual generic methods. former category, images or features generated benefit model. latter one, additional multi-sample sets used. There three types methods: combining features, improving loss function, network structure, all which covered our analysis. Moreover, review datasets that commonly used evaluating models go compare results different Additionally, discuss problems with existing methods, including information preservation domain adaption Furthermore, regard developing unsupervised promising future direction, point out semantic gap important issue needs be further considered.