All Courses
6 lessons

Diffusion Models

Generate images by learning to reverse noise

Lessons

  1. 01

    Denoising Intuition

    Learn to reverse a staircase of Gaussian noise.

    Easy
  2. 02

    Forward & Reverse Diffusion

    The two Markov chains that define the generative process.

    Hard
  3. 03

    U-Net Architecture

    Skip connections across a contracting-expanding path.

    Medium
  4. 04

    DDPM from Scratch

    Train a tiny diffusion model on MNIST, end to end.

    Hard
  5. 05

    Classifier-Free Guidance

    Steer generation without an auxiliary classifier.

    Hard
  6. 06

    Latent Diffusion

    Why Stable Diffusion works in VAE latent space.

    Hard