Artem Ex Machina™: become AI Arts Artisan from the ground up

Name: Artem Ex Machina™: become AI Arts Artisan from the ground up
Rating: 4.3 (5 reviews)

Build and train Generative Adversarial Networks in Python. Effectively apply pre-trained models for generative art!

Created byAlexander Abdulkader Kheirallah, PhD

Last updated 4/2023

English

What you'll learn

Generate beautiful AI art with pre-trained models
Build your own Generative Adversarial Networks (GANs) from scratch
Train GAN on a selected set of images
Create pieces of AI art with your own models
Become AI art artisan (generative artist)
Understand how Neural Networks and Generative Adversarial Networks (GAN) work
Work in your own AI workshop locally (CPU training) or remotely (GPU training)
Start and manage AI painting project

Course content

6 sections • 41 lectures • 3h 48m total length

What will you get from the course?3:01
Learn to build and train generative adversarial networks from scratch using Python, explore pre-trained GAN models, and apply ethics and project management across six course modules.
Intro0:20
Begin your first GAN training!2:41
Begin your first gan training by building a portable, containerized environment with Docker, enabling consistent art generation across local and cloud gpu setups.
Load a Docker container4:39
Load the docker container from the star file, validate the image, create a local repository folder, and link a volume to prepare for compiling the image.
Compile a Docker container5:16
Compile a docker container from a prepared image, link a local volume to the work directory, and prepare a training environment accessible via a Jupyter notebook at localhost:8888.
Connect to a Docker container3:14
Run a simple GAN training procedure9:51
Examine the evolution of GAN6:00
Explore GAN training across epochs, tracking the discriminator's accuracy as the generator improves, and observe the progression from random noise to handwritten digit-like, artful images.
Switch the container on and off2:45
Master switching a Docker container on and off, using the stop button or closing the browser, then start from the container and access images locally.
Change the memory and CPU of your container3:07
Adjust memory and cpu allocations for your container to optimize training, balancing host resources and available ram. Set cpu cores and ram in resources, then apply and restart.
Train your first GAN challenge1:06
Train your first GAN solution4:52
Train your first GAN with the gun trainer, using specific paths for images and models, a subset of images representing number eight, and note how epoch choices affect discriminator accuracy.
Section quiz

Intro0:09
What is Deep Learning?11:28
Explore how artificial intelligence, machine learning, and deep learning differ, and learn the benefits and use cases of deep learning, including artwork generation.
Artificial Neural Networks and their anatomy9:18
Explore the anatomy of artificial neural networks, from input to output layers, through weighted sums with bias, transfer and activation functions, and contrast feedforward and recurrent architectures.
Training an Artificial Neural Network13:51
Explore how multilayer feedforward neural networks act as universal approximators, and learn how gradient descent and backpropagation adjust weights and biases using loss functions to train from data.
What is a Generative Adversarial Network?2:37
Discover how generative adversarial networks use a discriminator and a generator with random noise to create images, then backpropagate to update the generator while training on real and fake samples.
Mnist clothing GAN challenge0:26
Mnist clothing GAN solution2:12
Train a MNIST clothing GAN on a bag-focused subset to generate realistic bag images and explore generator-discriminator dynamics and model repositories.
Create pieces of AI Arts with your own models3:54
Load your trained generator and discriminator from the repository and use the included notebook to generate new images from random noise, noting how noise and latent dim variable affect outputs.
Section quiz

Intro0:07
Local CPU vs remote GPU3:51
Explore the CPU versus GPU, noting CPU's sequential processing and GPU's parallel compute, and compare local versus remote work modes for art projects using Google Colab.
Introduction to Google Colab3:59
Access Google Colab to connect to a remote machine with higher computational capabilities, run Jupyter-like notebooks for Python code, and choose free or paid GPU options to train GAN models.
Principles of AI arts project management4:14
Organize your ai arts project with a single visual folder and a dedicated repository for models, training images, and evolution plots, clone the repository, and resize training images for consistency.
Easily collect your custom training images1:29
Learn to collect a large image dataset by installing the Chrome extension download all images and batch download them into a single zipped folder for training.
Section quiz

Intro0:07
Unpacking basic GAN architecture28:40
From scratch handwritten digit GAN0:35
From scratch handwritten digit GAN: solution14:07
CNNs & complex GAN architecture23:01
Grayscale landscape GAN1:57
Grayscale landscape GAN: solution17:45
Colour paintings GAN7:14
Learn to train a color portraits GAN from scratch using Google Colab and GPU, clone a repository, upload portraits of humans, and adjust the generator's convolutional layers for color output.
Colour painting GAN: solution11:08
Discover how to train a color painting GAN on a GPU, adjust resize settings for color images, build a three-upscale generator with a three-filter output, and navigate training challenges.
Section quiz

Requirements

Basic familiarity with python programming
Ability to work in a Docker container will be a bonus

Description

Did you know that computers can generate pictorial art? Have you ever wondered how can they do that? Did you know that the art generated by Artificial Intelligence (AI) has been sold on auctions for thousands of $$$?

Welcome to Artem Ex Machina™! Here you will learn how to become Artificial Intelligence (AI) Artisan by building, training and applying Generative Artificial Networks (GANs) - deep learning networks behind AI generated art. What makes this course unique is the combination of three elements:

Building, training and using GANs from scratch
Applying pre-trained GANs for beautiful AI Arts
Starting, maintaining and managing AI Arts projects

By the end of this course you will, therefore, be able to:

Build and train GANs using your own set of training images
Create pieces of AI Arts with publicly available, FREE, and amazing quality pre-trained GANs models
Understand how Neural Networks, in general, and GANs, in particular, work
Effectively manage AI Arts projects

To get a taster on the kind of images you'll be able to generate by completing this course, have a look at the logo of this course, which has 6 examples of images generated with GANs :)

The course uses Python as a programming language upon which you'll be able to build, train and use your own GANs. Not only you’ll gain practical capabilities when it comes to developing and diagnosing these deep learning models but also you'll gain a proper understanding of theory behind them. Since managing and, potentially, sharing your working environment behind GANs is an important part of the workflow of the generative artist, an integral part of the course is provision of AI-Art-compatible Docker containers and teaching you how to effectively use them. Understanding how GANs work combined with the ability to build them, will give you a strong generative artisanship acumen, and hence strong competitive advantage in the world of Data Science. In addition to that you'll learn how to make an effective use of pre-trained models that were trained for thousands of hours, to create high quality art. Therefore, you'll be able to work on your own local computer as well as remote high performance computing service such as Google Colab using Graphic Processing Units!

The course is split into 6 sections culminating with a 3 capstone projects that you can add to your portfolio:

Introduction. Here you'll learn how to setup your local working environment PLUS how to train and diagnose a hand-written digits GAN
What is a Generative Adversarial Network? Here you'll learn how GANs work and how they are trained PLUS how to train and apply GAN of images of clothing
AI Arts projects. Here you'll learn how to work locally vs how to work remotely and principles of AI Arts project management PLUS how to efficiently download bulk of images for your GAN training
GAN architectures. Here you'll learn how to build and train GAN models from scratch to produce complex grayscale and colour images
GAN+CLIP. Here you'll learn how to use pre-trained models for beautiful AI Arts generation
Capstone projects. Here you'll have the chance to complete three capstone projects to consolidate what you've learnt in the course.

Start you Artem Ex Machina™ journey now!

Who this course is for:

Data scientists and analyst that want to be able to build and apply Generative Adversarial Network models
Data scientists and analyst that want to learn or enhance their Deep Learning skills
STEM professionals wanting to expand into emerging word of AI arts
Anyone who wants to learn how to start AI Arts workshop
Anyone interested in AI arts and that wants to understand it

Artem Ex Machina™: become AI Arts Artisan from the ground up

What you'll learn

Explore related topics

Course content

Introduction12 lectures • 47min

What is a Generative Adversarial Network?8 lectures • 44min

AI Arts projects5 lectures • 14min

GAN architectures9 lectures • 1hr 45min

GAN+CLIP2 lectures • 15min

Capstone projects5 lectures • 5min

Requirements

Description

Who this course is for: