Generating Novel Imagery with Generative Adversarial Networks (GANs)

Generative Adversarial Networks, or GANs, have been making waves in the AI community for their uncanny ability to generate strikingly realistic images. They’ve been used to create everything from stunningly realistic human faces to artworks that look like they’ve been painted by a master artist. Today, I will provide a comprehensive guide on what GANs are and how they work, supplemented with Python code snippets on a real world example for clarity. <h1>What Are GANs?</h1> Invented by Ian Goodfellow and his team in 2014 (<a href="https://papers.nips.cc/paper_files/paper/2014/file/5ca3e9b122f61f8f06494c97b1afccf3-Paper.pdf" rel="noopener ugc nofollow" target="_blank">paper</a>), GANs are a type of machine learning system that consists of two main components: a Generator and a Discriminator. These two networks work against each other (thus the term “adversarial”), and this unique dynamic is what makes GANs so effective. As the generator and discriminator networks compete against each other, they both become better at their respective tasks. The generator learns to create more realistic images, and the discriminator learns to become more accurate at identifying fake images. <a href="https://medium.com/@mansuri.uzair/generating-novel-imagery-with-generative-adversarial-networks-gans-31cb76333891">Read More</a>