At a Glance

Name: Creating anime characters using DCGANs and PyTorch
Price: Free CAD
Rating: 4.6 (81 reviews)
Author: joseph_santarcangelo, artem_arutyunov, roodra_kanwar

Mass production of millions of unique anime characters is nearly impossible for even the most skilled painter, but it becomes feasible with the use of machine learning methods. In this guided project, you will have the opportunity to build machine learning models and generate anime characters on your own. Furthermore, you will explore the Deep Convolutional Generative Adversarial Networks (DCGANs) method, which is specifically designed for large-scale anime production.

About

You have been hired by a video game company as a data scientist to address the challenges they are facing and help save their business.

The company's game is known for its unique characters, customized for each player. However, as the player base has grown exponentially, it has become nearly impossible for the artists to manually create characters for millions of players. To retain their customers, your boss wants to find a solution that allows for mass production of anime characters using a machine-learning method.

As a data scientist, you are aware of the potential of Generative Adversarial Networks (GANs) to assist in this task. GANs are a class of machine learning frameworks that can generate new and realistic images that appear authentic to human observers. By combining GANs with Convolutional Neural Networks (CNNs), the process of generating images can be further enhanced, resulting in what is known as Deep Convolutional Generative Adversarial Networks (DCGANs).

Your objective is to train a DCGAN model using existing character data in order to produce a large number of unique anime characters for the video game.

A Look at the Project Ahead

In this guided project, you will begin by learning the basics of GANs, using toy data to understand the roles of the generator and discriminator components.
In the second part of the project, you will train Deep Convolutional Generative Adversarial Networks (DCGANs) models to create anime characters.
By the end of the project, you will have the following capabilities:

Understanding the fundamentals of GANs
Implementing GANs on datasets
Knowing how to train DCGANs
Generating a large quantity of unique images using DCGANs
Understanding the impact of changing the input of the latent space on the generated images

What You'll Need

This guided project is suitable for intermediate learners in the field of Machine Learning and Data Science.
A basic understanding of Python for data science is recommended before starting this project.
We recommend using the IBM Skills Network Labs environment for this guided project. Everything you need to complete this project will be provided to you via the Skills Network Labs. The platform is best supported on current versions of Chrome, Edge, Firefox, Internet Explorer, or Safari.