Deep face recognition: A survey

Deep learning applies multiple processing layers to learn representations of data with levels feature extraction. This emerging technique has reshaped the research landscape face recognition (FR) since 2014, launched by breakthroughs DeepFace and DeepID. Since then, deep technique, characterized hierarchical architecture stitch together pixels into invariant representation, dramatically improved state-of-the-art performance fostered successful real-world applications. In this survey, we provide a comprehensive review recent developments on FR, covering broad topics algorithm designs, databases, protocols, application scenes. First, summarize different network architectures loss functions proposed in rapid evolution FR methods. Second, related methods are categorized two classes: "one-to-many augmentation" "many-to-one normalization". Then, compare commonly used databases for both model training evaluation. Third, miscellaneous scenes such as cross-factor, heterogenous, multiple-media industrial Finally, technical challenges several promising directions highlighted.