(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Face recognition among women wearing hijab using deep learning

Face recognition among women wearing hijab using deep learning / Amirul Arif Ab Rasid
Tujuan projek ini adalah untuk mewujudkan pengenalan wajah yang lengkap di kalangan wanita yang memakai tudung. Hijab menyembunyikan rambut dan ciri luaran lain dari kepala (seperti telinga). Oleh itu, ia mungkin mempunyai implikasi bagi cara di mana wajah sedemikian dilihat. Dengan pengenalan “Deep Learning” (DL), konsep “Convolutional Neural Network” (CNN) yang pernah menjadi idea dapat direalisasikan. Dalam projek ini, “MATLAB Deep Network Designer” digunakan sebagai enjin teras untuk memberi kuasa pengenalan muka yang bertujuan membantu pengguna untuk mengenal pasti wanita yang memakai hijab. Pengguna boleh mengenal pasti wajah walaupun mereka memakai tudung. Pengenalan wajah juga boleh dilakukan dengan baik dalam mengenal pasti wajah-wajah berhadapan dengan betul dan tidak betul. Pengoptimuman dilakukan dengan bereksperimen secara berperingkat dengan beberapa parameter latihan untuk mendapatkan nilai terbaik untuk tujuan unik ini. Dengan menggunakan 3 seni bina model CNN yang terlatih iaitu AlexNet, GoogleNet dan Vgg16, prestasi terbaik algoritma latihan dapat dihasilkan untuk mengenali wajah. Pengoptimasi berdasarkan gabungan algoritma memainkan peranan penting dalam mengoptimumkan algoritma latihan. Perbandingan antara 3 model akan digambarkan dalam projek ini berdasarkan ketepatannya. Keputusan diperolehi untuk AlexNet (61%), GoogleNet (71%), dan Vgg16 (79%). _______________________________________________________________________________________________________ The aim of this project is to create a complete face recognition among ladies who wearing hijab. The hijab conceals hair and other external features of a head (such as the ears). It, therefore, may have implications for the way in which such faces are perceived. With the introduction of Deep Learning (DL), the concept of Convolutional Neural Network (CNN) that was once an idea can be realized. In this project, MATLAB Deep Network Designer is used as the core engine to power the face recognition that aims to help users to identify the faces of women wearing hijab. The users can identify the face even though they wearing hijab. The face recognition also can be performed well in identifying properly and improperly frontalized faces. Optimization is performed by experimenting in stages with several training parameters to obtain the best value for this unique purpose. By using 3 pre-trained CNN models architecture which is AlexNet, GoogleNet, and Vgg16, the best performance of the training algorithm can be produced in order to recognize the face. Combined-algorithm based optimizers play an important role in optimizing the training algorithm. The comparison between the 3 models will be illustrated in this project based on accuracy. The result obtained for AlexNet (61%), GoogleNet (71%), and Vgg16 (79%).
Contributor(s):
Amirul Arif Ab Rasid - Author
Primary Item Type:
Final Year Project
Identifiers:
Accession Number : 875008658
Language:
English
Subject Keywords:
face; recognition; hijab
First presented to the public:
6/1/2019
Original Publication Date:
3/5/2020
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 76
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2020-03-05 16:12:54.517
Date Last Updated
2020-12-02 15:33:01.583
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Face recognition among women wearing hijab using deep learning1 2020-03-05 16:12:54.517