Overview / Usage

The aim of this project is to create Art Generating AI based on the popular Deep Convolutional Generative Adversarial Networks with changes to account for novelty. This "painter" will be able to creatively come up with unique artworks that will look human-made. Such an application could have vast applications in the fields of game-development and possibly Virtual Reality.

Methodology / Approach

The modified DCGAN will learn from a sizeable dataset of artworks from Human history and adopt an approach based on psychological analysis of what makes art so "creative". I will be using Tensorflow and working on previously performed research in the field of creative AI with advancements to achieve best results. A possible extension into tweaked Wasserstein GANs will also be looked into.

Technologies Used

Intel Distribution for Python, TensorFlow* Optimizations for the Intel Architecture