NeurGAN Lab

A real GAN running in your browser. No APIs. No cloud. Just math.

Training Data

0 images
+

Drop images here or click to upload

Min 5 recommended · Images resized to 32×32

Architecture

z ~ N(0,1)
→
Generator
→
Fake Image
→
Discriminator
→
Real/Fake
G: Dense → Conv2DTranspose ×2 → Conv2D(tanh) | D: Conv2D ×3 → Dense(sigmoid)

Hyperparameters

Latent Dim64
Learning Rate0.0002
Batch Size8
Epochs500

Training

Ready to train
G Loss
0.0000
D Loss
0.0000
ms/step
0

Loss

GenDisc