Generative Deep Learning I

Teaching Machines To Generate, Edit, and Paint Images

Created by: Madali Nabil
Phd Student

What Will I Learn?

• Learn how images are encoded into numerical representations and decoded back into visual formats.
• Master techniques to edit images by manipulating their latent representations, enabling transformative changes and enhancements.
• Dive into GANs, including their structure, training dynamics with a generator and discriminator, and applications in generating realistic images.
• Examine advanced GAN architectures such as WGAN, Progressive GAN, StyleGAN v1, and StyleGAN v2, focusing on their improvements in stability and image quality.

Requirements

• Programming Skills: Proficiency in programming, preferably in Python, as the course involves hands-on implementation of deep learning models using libraries such as TensorFlow or PyTorch.
• Basic Knowledge of Deep Learning: Some prior exposure to deep learning concepts and architectures (e.g., feedforward neural networks, convolutional neural networks) will aid in grasping advanced topics covered in the course.
• Computer Vision Basics (Recommended): While not mandatory, familiarity with computer vision tasks and techniques can facilitate understanding the applications and challenges specific to image generation and manipulation.

Target audience

• Data Scientists and Machine Learning Engineers: Professionals looking to deepen their expertise in deep learning techniques for image generation and manipulation.
• Researchers and Academics: Those involved in academic research or industrial research labs focusing on computer vision, generative models, and AI.
• Computer Vision Enthusiasts: Individuals passionate about exploring cutting-edge technologies in computer vision and applying them to create realistic images.
• Entrepreneurs and Innovators: Individuals aiming to innovate in fields requiring advanced image manipulation capabilities, such as entertainment, fashion, healthcare, and more.