DeepLearning.AI Generative Adversarial Networks (GANs) Specialization

Welcome to the DeepLearning.AI Generative Adversarial Networks (GANs) Specialization repository! This repository contains all my coursework, assignments, and projects completed during the GANs Specialization program offered by DeepLearning.AI on Coursera.

Instructor

• Sharon Zhou

Syllabus

Course 1: Build Basic Generative Adversarial Networks (GANs)

In this first course, you’ll dive into the fundamentals of GANs, learning about their architecture and building your own GANs from scratch.

Week 1: Intro to GANs

• Learn about GANs and their applications.
• Understand the basic components of GANs and how they interact.
• Build your first GAN using PyTorch.

Week 2: Deep Convolutional GAN

• Implement convolutional layers to enhance GAN performance.
• Learn about activation functions, batch normalization, and transposed convolutions.
• Build an advanced DCGAN for image processing.

Week 3: Wasserstein GANs with Normalization

• Discover techniques like WGANs to address GAN training instability and mode collapse.
• Learn about W-Loss and Lipschitz Continuity.

Week 4: Conditional and Controllable GANs

• Build conditional GANs to generate specific categories of outputs.
• Explore how to control GAN outputs by modifying feature conditions.

Course 2: Build Better Generative Adversarial Networks (GANs)

This course focuses on evaluating and enhancing GAN performance, exploring the state-of-the-art advancements in GANs.

Week 1: GAN Evaluation

• Understand the challenges in evaluating GANs.
• Implement the Fréchet Inception Distance (FID) method to assess GAN accuracy and diversity.

Week 2: GAN Disadvantages and Bias

• Learn the disadvantages of GANs compared to other generative models.
• Explore sources of bias in GANs and techniques for detecting them.

Week 3: StyleGAN and Advancements

• Discover how StyleGAN advances previous models and implement key components of StyleGAN.

Course 3: Apply Generative Adversarial Networks (GANs)

This course covers the practical applications of GANs, focusing on data augmentation, privacy, and image-to-image translation.

Week 1: GANs for Data Augmentation and Privacy Preservation

• Explore GAN applications for data augmentation, privacy, and anonymity.
• Learn how GAN-generated data can enhance AI models.

Week 2: Image-to-Image Translation

• Implement Pix2Pix for paired image-to-image translation, adapting satellite images to map routes and vice versa.
• Explore the U-Net generator and PatchGAN discriminator architectures.

Week 3: Image-to-Image Unpaired Translation

• Implement CycleGAN for unpaired image-to-image translation, transforming horses into zebras and vice versa using two GANs in one.

Acknowledgments

Special thanks to Sharon Zhou and the DeepLearning.AI team for creating this comprehensive specialization that dives deep into the world of GANs.

---

Note: This repository is for educational purposes only.

---