General Adversarial Network for Transferring Car-Styles, machine learning,deep learning
In recent years, there is dramatic progress in machine learning fields, especially in the field of
generative models. Almost from every aspect, there are a lot of various flavors of generative
models. However, to be more specific in the point, Generative Adversarial Networks, in short GANs,
has seen spectacular consideration and advancements in deep learning research areas.
One application of GANs is the style transfer GAN. The aim of style transfer GAN is to
estimate the prospective distribution on real data samples and the artistic style within the images. In
addition, style transfer GAN is able to duplicate the same style for new - out of sample -
In this project work, we will design, implement and verify a style transfer GAN model that
is specifically customized for The Comprehensive Cars (CompCars) dataset. We will go further on
researching and developing models related to deep learning GANs Style Transfer.
A large number of types of generative Adversarial Networks have been proposed in recent years.
They are already proven to be really successful in many image and vision computing assignments
and one of them is is Style Transfer. Currently, Style Transfer methods have already seen
an incredible amount of research focus and a number of domains, it has been applied in
applications. The expected output is an application that is capable of generating, styling, and
customizing car vehicles.
Tasks to be performed by the student will include:
Collecting and running the relevant dataset and models
Designing a special GAN for
toy data sets
Validating that GAN on CompCars datasets
Identifying the main drawbacks
Training the data set
Building an application GUI with Angular Python
Improving the results by trying well-known methods
gépi tanulás, machine learning, deep learning, GANs