A step-by-step Colab notebook that shows how to generate, steer, and invert handwritten digits with three progressively richer models:
-
Unconditional DC-GAN – learns the data distribution and produces random digits.
-
Conditional GAN (cGAN) – adds class labels so you can ask for any specific numeral (0-9).
-
Encoder / Inference Net – maps a real image back to its latent vector z, enabling reconstructions and latent-space editing.