GAN Playground

Generative Adversarial Network — watch the generator vs discriminator battle

Step 0

Data Controls

Customize the training data and parameters

220

Generator

Noise → Fake Image

G
Noise z
Fake Image
Generator Loss2.500

Discriminator

Real / Fake Classification

D
Fake
Real
Hidden
Hidden
Prediction
50%
Fake
Discriminator Loss1.200

Training Progress

Generator vs Discriminator loss over time

Step 0
Press “Auto Train” or “Step” to begin

Image Evolution

How the generated image changes over training steps

Live