Document Type : Original full papers (regular papers)
Authors
1
Department of Computer Science, Faculty of Computers and Artificial Intelligence, Fayoum University, Fayoum, Egypt
2
Department of Computer Science, Faculty of Computers and Artificial Intelligence, Fayoum University, Fayoum, Egypt,
3
Department of Computer Science, Faculty of Computers and Information, Fayoum University, Fayoum, Egypt
Abstract
The humans' faces hold a lake of information that enables us to identify them. Classification of a human's ethnicity is important information in various areas, such as biometrics, security, and personal safety. Physical characteristics, such as skin color, hair type, and facial features are used by the human brain to divide people into different ethnic groups. This paper presents a model that classifies ethnicity and gender according to 5 ethnicity classes. Moreover, the proposed model classifies males and females for each ethnicity group. The proposed model uses deep learning to mimic the behavior of the human brain in distinguishing between different ethnic groups. The proposed model comprises two main phases, the first phase is the data preparation and preprocessing which is a crucial step to make the facial images meet the requirements of the second phase and emphasize that the proposed model is more robust and generalized, this phase includes three main steps; Data Augmentation by using images flipping and noise injection techniques, Face Detection, and Images Resizing, the second phase is features extraction and classification with convolutional neural network (CNN). The proposed model gives promising results with accuracy 99.78%, 100% for ethnicity and gender, respectively.
Keywords
Main Subjects
)
View on SCiNiTO