VizuaraVizuara AI Pods

Principles of Diffusion Models

From variational autoencoders to video generation — understanding diffusion models from first principles.

intermediate~4 hours6 pods live

Pods in this Course

Denoising Diffusion Probabilistic Models (DDPM)
1

Denoising Diffusion Probabilistic Models (DDPM)

A step-by-step guide to understanding how DDPMs learn to generate images by mastering the art of denoising.

~4h4 notebooksCase study
Energy-Based Models and the Score Function
2

Energy-Based Models and the Score Function

How an energy landscape, a gradient, and a drunk hiker unlock modern generative modeling.

~4h4 notebooksCase study
Denoising Score Matching
3

Denoising Score Matching

How a simple trick of adding noise solves the hardest problem in score-based generative models.

~3h3 notebooksCase study
Noise Conditioned Score Networks (NCSN)
4

Noise Conditioned Score Networks (NCSN)

How Song & Ermon solved the low-density problem in score matching by conditioning on multiple noise levels.

~4h4 notebooksCase study
From Still to Motion: How Diffusion Models Learned to Generate Videos
5

From Still to Motion: How Diffusion Models Learned to Generate Videos

Extending image diffusion to the temporal dimension — from 2D noise to coherent video, one frame at a time.

~4h3 notebooksCase study
Variational Autoencoders From Scratch
6

Variational Autoencoders From Scratch

Building VAEs from first principles — intuition, math, and a full PyTorch implementation.

~4h4 notebooksCase study