Understanding Deepfakes with Keras.
This is a project-based course which should take approximately 2 hours to finish. Before diving into the project, please take a look at the course objectives and structure:
Course Objectives
In this course, we are going to focus on two learning objectives:
Implement a Deep Convolutional Generative Adversarial Network (DCGAN).
Train a DCGAN to synthesize realistic looking images.
By the end of this course, you will understand how to implement DCGANs, and how to train them to generate realistic synthetic images.
Course Structure
This course is divided into 3 parts:
Course Overview: This introductory reading material.
Understanding Deepfakes with Keras: This is the hands on project that we will work on in Rhyme.
Graded Quiz: This is the final assignment that you need to pass in order to finish the course successfully.
Project Structure
The hands on project on Understanding Deepfakes with Keras is divided into following tasks:
Task 1: Introduction
Introduction to the problem.
Introduction to the Rhyme interface.
Importing required libraries and helper functions.
Task 2: Importing and Plotting the Data
Importing the MNIST Dataset
Creating a subset of the dataset for just one class.
Visualizing the subset.
Task 3: Discriminator
Basic understanding of how a GAN works.
Creating a Discriminator Network.
Creating an optimizer instance.
Task 4: Generator
Creating a Generator Network.
Generating a new image from the untrained Generator model.
Task 5: Generative Adversarial Network (GAN)
Connecting the Generator and Discriminator to create a Generative Adversarial Network (GAN)
Task 6: Training the GAN
Creating a training loop.
Creating a dynamic plot that displays generated images after each epoch.
Task 7: Final Results
Understanding the final results.
