FACE VERIFICATION WITH STATISTICAL MODELS OF SHAPE AND APPEARANCE
Need help with a related project topic or New topic? Send Us Your Topic
DOWNLOAD THE COMPLETE PROJECT MATERIAL
FACE VERIFICATION WITH STATISTICAL MODELS OF SHAPE AND APPEARANCE
Chapter One: Introduction
Computer vision and machine learning research are major areas of study in top computer science departments across the world. It has resulted in numerous achievements in both academic research and commercial applications.
Face recognition is a hot topic in computer vision and machine learning research right now. It was described in [30] as one of the most effective applications of image analysis and comprehension, with two reasons cited for the considerable research efforts in this area: the vast range of applications it provides and the availability of technology to assist the research.
A review of some of the leading institutions in this and other fields of computer science finds that each has a robust code repository that has been constructed over time by researchers and is available for new researchers to build on, hence speeding up research activity.
Examples include VisionX from Cornell University’s Vision and Image Analysis Group1, FSL from the Analysis Group at FMRIB in Oxford, UK2, and STAIR Vision Library (SVL)3, which was built by a Stanford PhD student for research purposes, first to help the Stanford AI robot project.
The computer vision and machine learning (CVML) group at The Africa University of Science and Technology, Abuja, aims to create its own code repository from the ground up in order to ease group research.
Many of the algorithms developed for the repository are not publicly available elsewhere, or at least not in the organised manner that was used in the repository.
Our primary purpose for this thesis was to create a code repository for the computer vision and machine learning group, which focuses on research in face recognition and computer animation, as well as to use the code base to conduct face verification tests.
The tests investigated the use of Mahalanobis distance, Euclidean distance, Manhattan distance, and normalised correlation as metrics for face verification, based on appearance model parameters (2). These measurements were chosen because they are the most widely used.
This chapter briefly covers the code repository’s initial design, the scope of this thesis, our contributions, and the report’s presentation.
1.1 Scope: The image below depicts the suggested framework for the code repository used in this thesis work.
Figure 1.1: CVML Code Repository Initial Framework
3
These phases summarise our thesis tasks.
1. Create the following models.
a. Texture Model b. Appearance Model (APM) c. Active Shape Model (ASM) 2. Experiment with both ASM and APM for facial verification.
1.2 Contributions.
The following contributions were made towards the purpose of this thesis.
1. Created code for the texture model.
2. Created code for the Appearance model.
3. Contributed code for creating Active Shape Model.
4. Used the repository code to conduct verification experiments with ASMs and APMs.
1.3 Layout.
The remaining chapters of this thesis are organised as follows:
Chapter 2 addresses the literature on certain foundational concepts required for a full comprehension of the thesis work, as well as some of the mathematical techniques employed and several approaches to face detection and recognition.
Chapter 3 explains the steps needed in creating the active shape model.
Chapter 4 describes how to create appearance models and utilise an ASM to understand images.
Chapter 5 describes our face verification experiment framework.
Chapter 6 summarises the experiments and the outcomes.
Chapter 7 summarises the thesis’ achievements, problems, and future work.
Need help with a related project topic or New topic? Send Us Your Topic