ETRO VUB
About ETRO  |  News  |  Events  |  Vacancies  |  Contact  
Home Research Education Industry Publications About ETRO

Master theses

Current and past ideas and concepts for Master Theses.

Deep Generative Models for Pose-Invariant Face Recognition

Subject

For almost half of a century, face recognition has been one of the most intensively studied topics in computer vision thanks to its applications in biometric authentication, surveillance, etc. With a recent shift in the mobile industry from using fingerprint and iris scanning to face scanning for unlocking devices, the topic is gaining even more attention. Although many great achievements have been made, one of the major challenges still lies in the fact that the appearance change from pose variation of a person is often more significant than the intrinsic differences between individuals. To overcome the problem, current research generally follows two main approaches: either extracting discriminative pose-robust features as face representations or synthesizing face images from one pose to another pose. In this Master’s thesis, we will explore the possibilities of applying recent advancement in deep learning and Generative Adversarial Networks (GANs) to both ideas above.

Related articles:
• https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/pose_face_recognition.pdf
• http://cvlab.cse.msu.edu/pdfs/Tran_Yin_Liu_CVPR2017.pdf

Kind of work

The thesis will consist of the following steps:
1. Review the literature in pose-invariant face recognition
2. Select a dataset or build a new one
3. Re-implement state-of-the-art algorithms, with the focus on deep-learning and GANs methods (computing resource is supported if necessary)
4. Analyse the strong points and drawbacks of each method
5. Improve the existing algorithms
6. Write the thesis 

Framework of the Thesis

Big heterogeneous data processing and analytics is an active research direction at VUB-ETRO with research focusing on centralized, distributed and embedded algorithms for highly practical applications. The work within this thesis will go fully in line with research in these themes at the lab. Publication of the research results in international conferences and journals is also encouraged and supported.

Number of Students

1

Expected Student Profile

• Interest/knowledge in Machine Learning
• Good programming skills in Python and/or MATLAB. Experience with C/C++ and Linux is a plus.
• Good understanding of Mathematics (Linear Algebra, Calculus and Optimization).

Promotor

Prof. Dr. Ir. Nikolaos Deligiannis

+32 (0)2 629 1683

ndeligia@etrovub.be

more info

Supervisor

Mr. Hung Le

+32 (0)2 629 2930

dle@etrovub.be

more info

- Contact person

- IRIS

- AVSP

- LAMI

- Contact person

- Thesis proposals

- ETRO Courses

- Contact person

- Spin-offs

- Know How

- Journals

- Conferences

- Books

- Vacancies

- News

- Events

- Press

Contact

ETRO Department

info@etro.vub.ac.be

Tel: +32 2 629 29 30

©2019 • Vrije Universiteit Brussel • ETRO Dept. • Pleinlaan 2 • 1050 Brussels • Tel: +32 2 629 2930 (secretariat) • Fax: +32 2 629 2883 • WebmasterDisclaimer