|
Document Type |
: |
Thesis |
|
Document Title |
: |
Masked Face Recognition Based on Convolutional Neural Network and Using Combination of Synthetic and Realistic Masked Faces Datasets. التعرف على الوجوه المقنعة باستخدام نهج الشبكة العصبية التلافيفية واستخدام مجموعة من قواعد البيانات لصور وجوه مقنعة بطريقة واقعية واصطناعية |
|
Subject |
: |
Faculty of Computing and Information Technology |
|
Document Language |
: |
Arabic |
|
Abstract |
: |
In the last decade, face recognition models achieved high performance using deep learning approaches. Some recent models outperform or are close to the human ability to recognize and verify identities from human face images. However, deep learning- based models can still be deceived more quickly than humans due to different challenges, including illumination, poses, expressions, and occlusion. Masks are considered a partial occlusion challenge. So, it decreases the performance of face recognition models trained on unmasked faces. Over time, pandemics have recurred, such as the recent COVID-19 era. Such a disease forces people to wear masks to reduce the spread of the virus, which made it necessary to improve the performance of masking face recognition models, particularly in environments that are sensitive to security, such as airports and government facilities. This work aims to enhance masked face recognition (MFR) models using a convolutional neural network (CNN) and a combination of realistic and synthetic masked face image datasets. The used datasets were studied carefully and chosen based on critical effecting factors, which depend on training the model on images that have been taken in the circumstances close to real life and cover more challenges to enhance the quality and performance of models. FaceNet and Inception-ResNet-V1 were trained, customized, and evaluated based on three scenarios. These scenarios include alternating the training, validation, and testing between masked and unmasked face datasets. As a result, we outperform other studies that use similar datasets on masked face recognition. |
|
Supervisor |
: |
Dr. Salma Mohamad Kammoun |
|
Thesis Type |
: |
Master Thesis |
|
Publishing Year |
: |
1444 AH
2023 AD |
|
Co-Supervisor |
: |
Dr. Faris Anwar Kateb |
|
Added Date |
: |
Friday, May 5, 2023 |
|
Researchers
| نهى خالد بابكر | Babakr, Nuha Khaled | Researcher | Master | |
|