What you will learn
- Understand the basic intuition about GANs
- Generate images of digits (0 – 9) using DCGAN and WGAN
- Transform satellite images into maps using Pix2Pix architecture
- Transform zebras into horses using CycleGAN architecture
- Transfer styles between images
- Apply super resolution to improve image quality using ESRGAN architecture
- Create new faces of people with high quality and definition using StyleGAN
- Generate images through textual descriptions
- Restore old photos using GFP-GAN
- Complete missing parts of images using Boundless architecture
- Generate deepfakes to swap faces with SimSwap
Requirements
- Programming logic
- Basic Python programming
- Knowledge about neural networks is desirable, but not mandatory
Description
GANs (Generative Adversarial Networks) are considered one of the most modern and fascinating technologies within the field of Deep Learning and Computer Vision. They have gained a lot of attention because they can create fake content. One of the most classic examples is the creation of people who do not exist in the real world to be used to broadcast television programs. This technology is considered a revolution in the field of Artificial Intelligence for producing high quality results, remaining one of the most popular and relevant topics.
In this course you will learn the basic intuition and mainly the practical implementation of the most modern architectures of Generative Adversarial Networks! This course is considered a complete guide because it presents everything from the most basic concepts to the most modern and advanced techniques, so that in the end you will have all the necessary tools to build your own projects! See below some of the projects that you are going to implement step by step:
- Creating of digits from 0 to 9
- Transforming satellite images into map images, like Google Maps style
- Convert drawings into high-quality photos
- Create zebras using horse images
- Transfer styles between images using paintings by famous artists such as Van Gogh, Cezanne and Ukiyo-e
- Increase the resolution of low quality images (super resolution)
- Generate deepfakes (fake faces) with high quality
- Create images through textual descriptions
- Restore old photos
- Complete missing parts of images
- Swap the faces of people who are in different environments
To implement the projects, you will learn several different architectures of GANs, such as: DCGAN (Deep Convolutional Generative Adversarial Network), WGAN (Wassertein GAN), WGAN-GP (Wassertein GAN-Gradient Penalty), cGAN (conditional GAN), Pix2Pix (Image-to-Image), CycleGAN (Cycle-Consistent Adversarial Network), SRGAN (Super Resolution GAN), ESRGAN (Enhanced Super Resolution GAN), StyleGAN (Style-Based Generator Architecture for GANs), VQ-GAN (Vector Quantized Generative Adversarial Network), CLIP (Contrastive Language–Image Pre-training), BigGAN, GFP-GAN (Generative Facial Prior GAN), Unlimited GAN (Boundless) and SimSwap (Simple Swap).
During the course, we will use the Python programming language and Google Colab online, so you do not have to worry about installing and configuring libraries on your own machine! More than 100 lectures and 16 hours of videos!
Who this course is for
- People interested in creating complex applications using GANs
- Undergraduate and graduate students who are taking courses on Computer Vision, Artificial Intelligence, Digital Image Processing or Computer Vision
- People who want to implement their own projects using Computer Vision techniques
- Data Scientists who want to increase their project portfolio
Love this !!!