Ethnic classification of face images using inductive learning

Dissertant

al-Hiti, Khaldun Abd Allah

Thesis advisor

Abu Suud, Salih M.

Comitee Members

al-Qaddumi, Ashraf Ahmad
Salamah, Walid
al-Dawud, Ali

University

Princess Sumaya University for Technology

Faculty

King Hussein Faculty for Computing Sciences

Department

Department of Computer Sciences

University Country

Jordan

Degree

Master

Degree Date

2014

English Abstract

This thesis provides the framework of new research in regards to racial classification of human race, taking into consideration Inductive Learning Algorithm (ILA) one of the machine learning algorithms. Experiments show that this approach has significant importance and results comparable to other approaches as the use of image treatments or neural networks. The focus of the thesis is divided into two phases: A - To sort out racial classification into three races, (Arab, Asian and Caucasian) in the evaluation part of the thesis compared with our work will be proved to complete previous work as results that used neural networks. B- Negro race was added as a new race , which has many various unique characteristics compared with other race such as lips and nose. The mechanism of both phases were the same, implementing the same collection of human faces images have been taken in previous research using neural network (NN), and pictures of (40 Arabic, 43 Asian and 67 Caucasian) in addition to 70 pictures of Negro race. The methodology of the experiment is based on submitting all faces details manually into Excel spreadsheet, each column contains certain characteristics that refer to face detail as a question or drop down list of options, thereafter, second part of this experiment was inserting these data into the inductive learning algorithm (ILA). The used dataset is divided into a training set and test (unseen) set, the main objective of training set is to build a model (generated rules), while the test set is to validate the built model.

The remaining part of the dataset was used for the test.

Obtained result from the first phase of three races was 86.36 percent, which was more accurate than the previous results of neural networks that was 83.5 percent. But after the addition of another race (black), the result was slightly less accurate than the previous result, which was 82.63 percent, and this shows the power and great ability of (ILA) in classification human faces through both phases.

Main Subjects

Information Technology and Computer Science

Topics

• Inductive logic

No. of Pages

120

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-414241