Adaptive Hyperparameter for Face Recognition
Thanh-Tam Nguyen1, Son-Thai LE2, Van-Thuy LE3

1Thanh-Tam NGUYEN*, Faculty of Multimedia, The Posts and Telecommunications Institute of Technologies, Hanoi, Vietnam.
2Son-Thai LE, Department of Multimedia Communications, The School of Information and Communication Technology, Thai Nguyen University, Thai Nguyen, Vietnam.
3Van-Thuy LE, the School of Foreign Languages  Thai Nguyen University, Thai Nguyen, Vietnam. 

Manuscript received on December 19, 2020. | Revised Manuscript received on January 01, 2020. | Manuscript published on January 10, 2021. | PP: 116-119 | Volume-10 Issue-3, January 2021 | Retrieval Number: 100.1/ijitee.C84090110321| DOI: 10.35940/ijitee.C8409.0110321
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: One of the widely used prominent biometric techniques for identity authentication is Face Recognition. It plays an essential role in many areas, such as daily life, public security, finance, the military, and the smart school. The facial recognition task is identifying or verifying the identity of a person base on their face. The first step is face detection, which detects and locates human faces in images and videos. The face match process then finds an identity of the detected face. In recent years there have been many face recognition systems improving the performance based on deep learning models. Deep learning learns representations of the face based on multiple processing layers with multiple levels of feature extraction. This approach has made sufficient improvement in face recognition since 2014, launched by the breakthroughs of Deep Face and Deep ID. However, finding a way to choose the best hyperparameters remains an open question. In this paper, we introduce a method for adaptive hyperparameters selection to improve recognition accuracy. The proposed method achieves improvements on three datasets. 
Keywords: Face Recognition, Deep Learning, Hyperparameter.