AI Workshop: Hands-on with GANs with Deep Convolutional Networks

Learn how to build and train deep convolutional generative adversarial networks (GANs) in this interactive, workshop-style coding course.

Introduction

• A quick overview of GANs

1. Understanding Convolutional and Pooling Layers

• Course outline and prerequisites
• Setting up Google Colab cloud-hosted notebooks
• Understanding convolutional neural networks
• Transforming a multichannel image to tensor
• Applying convolutional and pooling layers
• Viewing the effect of different filters

3. Training a Discriminator as a Classification Model

• Types of convolutional layers
• Training data for discriminator bad fakes and real images
• Loading and transforming training image data
• Understanding the discriminator architecture
• Training a discriminator on bad fakes
• Training data for discriminator good fakes and real images
• Training a discriminator on good fakes

3. Training a Deep Convolutional GAN

• Generator and discriminator
• Deep convolutional GANs (DCGANs)
• Setting up data for GAN training
• Setting up the generator and discriminator
• Output from an untrained generator and discriminator
• Setting up the GAN training loop
• Viewing GAN training results

Conclusion

• Summary and next steps