Generative Adversarial Networks Cookbook电子书

Title Page

Copyright and Credits

Generative Adversarial Networks Cookbook

About Packt

Why subscribe?

Packt.com

Dedication

Contributors

About the author

About the reviewer

Packt is searching for authors like you

Preface

Who this book is for

What this book covers

To get the most out of this book

Download the example code files

Download the color images

Conventions used

Sections

Getting ready

How to do it…

How it works…

There's more…

See also

Get in touch

Reviews

Dedication2

What Is a Generative Adversarial Network?

Introduction

Generative and discriminative models

How to do it...

How it works...

A neural network love story

How to do it...

How it works...

Deep neural networks

How to do it...

How it works...

Architecture structure basics

How to do it...

How it works...

Basic building block – generator

How to do it...

How it works...

Basic building block – discriminator

How to do it...

How it works...

Basic building block – loss functions

How to do it...

How it works...

Training

How to do it...

How it works...

GAN pieces come together in different ways

How to do it...

How it works...

What does a GAN output?

How to do it...

How it works...

Working with limited data – style transfer

Dreaming new scenes – DCGAN

Enhancing simulated data – simGAN

Understanding the benefits of a GAN structure

How to do it...

How it works...

Exercise

Data First, Easy Environment, and Data Prep

Introduction

Is data that important?

Getting ready

How to do it...

How it works...

There's more...

But first, set up your development environment

Getting ready

How to do it...

Installing the NVIDIA driver for your GPU

Installing Nvidia-Docker

Purging all older versions of Docker

Adding package repositories

Installing NVIDIA-Docker2 and reloading the daemon

Testing nvidia-smi through the Docker container

Building a container for development

There's more...

Data types

Getting ready

How to do it...

How it works...

Running this code in the Docker container

There's more...

Data preprocessing

Getting ready

How to do it...

How it works...

There's more...

Anomalous data

Getting ready

How to do it...

Univariate method

There's more...

Balancing data

Getting ready

How to do it...

Sampling techniques

Random undersampling

Random oversampling

Synthetic minority oversampling technique

Ensemble techniques

Bagging

Boosting

AdaBoost

There's more...

Data augmentation

Getting ready

How to do it...

How it works...

There's more...

Exercise

My First GAN in Under 100 Lines

Introduction

From theory to code – a simple example

Getting ready

How to do it...

Discriminator base class

Generator base class

GAN base class

See also

Building a neural network in Keras and TensorFlow

Getting ready

How to do it...

Building the Docker containers

The Docker container

The run file

See also

Explaining your first GAN component – discriminator

Getting ready

How to do it...

Imports

Initialization variables (init in the Discriminator class)

Model definition for the discriminator

Helper methods in the Discriminator class

Explaining your second GAN component – generator

Getting ready

How to do it...

Imports

Generator initialization

Model definition of the generator

Helper methods of the generator

Putting all the GAN pieces together

Getting ready

How it works...

Step 1 – GAN class initialization

Step 2 – model definition

Step 3 – helper functions

Training your first GAN

Getting ready

How to do it...

Training class definition

Imports

init method in class

Load data method

Training method

Helper functions

Run script definition

Training the model and understanding the GAN output

Getting ready

How to do it...

How it works...

Exercise

Dreaming of New Outdoor Structures Using DCGAN

Introduction

What is DCGAN? A simple pseudocode example

Getting ready

How to do it...

Generator

Discriminator

See also

Tools – do I need any unique tools?

Getting ready

How to do it...

The development environment for DCGAN

Downloading and unpacking LSUN data

There's more...

See also

Parsing the data – is our data unique?

Getting ready

How to do it...

Code implementation – generator

Getting ready

How to do it...

Initializing generator – the DCGAN update

Building the DCGAN structure

See also

Code implementation – discriminator

Getting ready

How to do it...

Initializing the Discriminator class

Building the model structure

See also

Training

Getting ready

How to do it...

Changes to class initialization

Understanding the changes in pseudocode

The new and improved training script

Python run script

Shell run script

Evaluation – how do we know it worked?

Getting ready

How it works...

Adjusting parameters for better performance

How to do it...

Training parameters

Discriminator and generator architecture parameters

Exercise

Pix2Pix Image-to-Image Translation

Introduction

Introducing Pix2Pix with pseudocode

Getting ready

How to do it...

Discriminator

Generator

Parsing our dataset

Getting ready

How to do it...

Building the Docker container with a new Dockerfile

Building the auxiliary scripts

Code implementation – generator

Getting ready

How to do it...

Code – the GAN network

Getting ready

How to do it...

Code implementation – discriminator

Getting ready

How it works...

Training

Getting ready

How to do it...

Setting up the class

Training method

Plotting the results

Helper functions

Running the Training Script

Exercise

Style Transfering Your Image Using CycleGAN

Introduction

Pseudocode – how does it work?

Getting ready

How to do it...

What is so powerful about CycleGAN?

Parsing the CycleGAN dataset

Getting ready

How to do it...

Docker implementation

The data download script

What does the data actually look like?

Code implementation – generator

Getting ready

How to do it....

Code implementation – discriminator

Getting ready

How to do it...

Code implementation – GAN

Getting ready

How to do it...

On to training

Getting ready

How to do it...

Initialization

Training method

Helper method

Exercise

Using Simulated Images To Create Photo-Realistic Eyeballs with SimGAN

Introduction

How SimGAN architecture works

Getting ready

How to do it...

Pseudocode – how does it work?

Getting ready

How to do it...

How to work with training data

Getting ready

How to do it...

Kaggle and its API

Building the Docker image

Running the Docker image

Code implementation – loss functions

Getting ready

How to do it...

Code implementation – generator

Getting ready

How to do it...

Boilerplate items

Model development

Helper functions

Code implementation – discriminator

Getting ready

How to do it...

Boilerplate

Model architecture

Helper functions

Code implementation – GAN

Getting ready

How to do it...

Training the simGAN network

Getting ready

How to do it...

Initialization

Training function

Helper functions

Python run script

Shell run script

Exercise

From Image to 3D Models Using GANs

Introduction

Introduction to using GANs in order to produce 3D models

Getting ready

How to do it...

For a 2D image – learning an encoding space for an image

Training a model using 3D convolutions

Environment preparation

Getting ready

How to do it...

Creating the Docker container

Building the Docker container

Encoding 2D data and matching to 3D objects

Getting ready

How to do it...

Code to run a simple encoder

The shell script to run the encoder with our Docker container

Code implementation – generator

Getting ready

How to do it...

Generator class preparation

Building the generator model

Code implementation – discriminator

Getting ready

How to do it...

Discriminator class preparation

Building the discriminator model

Code implementation – GAN

Getting ready

How to do it...

Training this model

Getting ready

How to do it...

Training class preparation

Helper functions

The training method

Plotting the output of the network

Running the training script

Exercise

Other Books You May Enjoy

Leave a review - let other readers know what you think