Improving Open-set Face Recognition through Neural Ensemble, Maximal Entropy Loss, and Feature Augmentation

[Submitted on 23 Aug 2023]

Click here to download a PDF of the paper titled “Open-set Face Recognition with Neural Ensemble, Maximal Entropy Loss and Feature Augmentation” by Rafael Henrique Vareto, Manuel Günther, and William Robson Schwartz.

Download PDF

Abstract: Open-set face recognition involves biometric systems that may not have complete knowledge of all existing subjects. This means that they should be able to distinguish between registered and unregistered individuals. In this work, we propose a novel method that uses a combination of compact neural networks and a margin-based cost function to handle this challenge. We also introduce a feature augmentation approach that allows for the generation of additional negative samples. Our experiments on the LFW and IJB-C datasets demonstrate that our approach improves both closed and open-set identification rates.