Human Faces Generation using GAN
Project Description
A generative adversarial network (GAN) is a type of machine learning algorithm where two neural networks contest with each other. The first network generates candidates while the second network evaluates the candidates. Given a training set, this technique learns to generate new data with the same statistics as the training set.
For instance, a GAN trained on a set of face pictures can generate realistic, albeit completely new pictures that look authentic to human observers.
In this project, the student(s) will explore how to develop an architecture to identify facial expressions from existing images, and generate images embedding the same facial expressions.
For instance, a GAN trained on a set of face pictures can generate realistic, albeit completely new pictures that look authentic to human observers.
In this project, the student(s) will explore how to develop an architecture to identify facial expressions from existing images, and generate images embedding the same facial expressions.
Supervisor
HUI Pan
Quota
3
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP4100
Applicant's Roles
Required Skills: python, basics of machine learning frameworks (tensorflow, keras, pytorch), and computer vision techniques (haar cascades).
The applicant(s) will conduct research under the supervision of the professor and other senior postgraduate students in the research group. They will participate in the development and the evaluation of the proposed solution.
The applicant(s) will conduct research under the supervision of the professor and other senior postgraduate students in the research group. They will participate in the development and the evaluation of the proposed solution.
Applicant's Learning Objectives
By the end of the project, the student(s) will be familiar with the basics of GAN, and will know how to generate images using such techniques.
Complexity of the project
Moderate