Generative Adversarial Networks Projects : Build Next-Generation Generative Models Using TensorFlow and Keras.

Description

In this book, we will use different complexities of datasets in order to build end-to-end projects. With every chapter, the level of complexity and operations will become advanced. It consists of 8 full-fledged projects covering approaches such as 3D-GAN, Age-cGAN, DCGAN, SRGAN, StackGAN, and CycleGAN with real-world use cases. Cover Title Page Copyright and Credits About Packt Contributors Table of Contents Preface Chapter 1: Introduction to Generative Adversarial Networks What is a GAN? What is a generator network? What is a discriminator network? Training through adversarial play in GANs Practical applications of GANs The detailed architecture of a GAN The architecture of the generator The architecture of the discriminator Important concepts related to GANs Kullback-Leibler divergence Jensen-Shannon divergence Nash equilibrium Objective functions Scoring algorithms The inception score The Fréchet inception distanceVariants of GANs Deep convolutional generative adversarial networks StackGANs CycleGANs 3D-GANs Age-cGANs pix2pix Advantages of GANs Problems with training GANs Mode collapse Vanishing gradients Internal covariate shift Solving stability problems when training GANs Feature matching Mini-batch discrimination Historical averaging One-sided label smoothing Batch normalization Instance normalization Summary Chapter 2: 3D-GAN -- Generating Shapes Using GANs Introduction to 3D-GANs 3D convolutions The architecture of a 3D-GAN The architecture of the generator networkThe architecture of the discriminator network Objective function Training 3D-GANs Setting up a project Preparing the data Download and extract the dataset Exploring the dataset What is a voxel? Loading and visualizing a 3D image Visualizing a 3D image A Keras implementation of a 3D-GAN The generator network The discriminator network Training a 3D-GAN Training the networks Saving the models Testing the models Visualizing losses Visualizing graphs Hyperparameter optimization Practical applications of 3D-GANs Summary Chapter 3: Face Aging Using Conditional GANIntroducing cGANs for face aging Understanding cGANs The architecture of the Age-cGAN The encoder network The generator network The discriminator network Face recognition network Stages of the Age-cGAN Conditional GAN training The training objective function Initial latent vector approximation Latent vector optimization Setting up the project Preparing the data Downloading the dataset Extracting the dataset A Keras implementation of an Age-cGAN The encoder network The generator network The discriminator network Training the cGAN Training the cGANInitial latent vector approximation Latent vector optimization Visualizing the losses Visualizing the graphs Practical applications of Age-cGAN Summary Chapter 4: Generating Anime Characters Using DCGANs Introducing to DCGANs Architectural details of a DCGAN Configuring the generator network Configuring the discriminator network Setting up the project Downloading and preparing the anime characters dataset Downloading the dataset Exploring the dataset Cropping and resizing images in the dataset Implementing a DCGAN using Keras Generator Discriminator