In this course, we will briefly overview some generative models that existed before GANs, go through the history of GANs and their applications to real-life problems, like image super-resolution, deblurring, and style transfer.
Syllabus:
1. Overview of different generative models and their issues
Hidden Markov Model. Restricted Boltzman Machine. Variational Autoencoders.
2. GAN and cGAN
A brief introduction to GAN idea. Vanilla convolutional GANs.
3. Application of GANs to real-life problems
Image super-resolution, deblurring and style transfer.
4. What is wrong with GANs? Modern approaches to GANs training and latest results.
Wasserstein GAN. Improved Training of Wasserstein GANs. What is wrong with Prisma app and how to fix it?
Course topics
Computer Vision, Deep Learning, GAN, Deep Generative Models.
Course tools
Pytorch or Tensorflow (not Keras)
Prerequisites
Having good programming skills and being familiar with Pytorch or Tensorflow (not Keras).
Lecturer
Oles DobosevychDeputy Dean at Ukrainian Catholic University / Computer Vision at DatAI
Oles is pursuing Ph.D. in Function Analysis and currently researching GANs applications. He is Deputy Dean of Faculty of Sciences and Computer Vision Engineer in DatAI. He has more than 8 years of experience (from Web development to Analytics), and during last three years, he has been consulting different IT-companies.
Fields of interests: Computer Vision, Deep Learning, Functional Analysis.
Contacts: dobosevych@ucu.edu.ua
Facebook: dobosevych, Skype: dobosevych
