In this paper. we propose a novel loss function for deep face recognition. called the additive orthant loss (Orthant loss). which can be combined for softmax-based loss functions to improve the feature-discriminative capability. https://www.lightemupsequences.com/