Skip to content

Chapter 27

Chapter 27 sections from Deep Learning with PyTorch.

deep-learning pytorch

Sections

  1. 27.1 Forward Diffusion Processes
  2. 27.2 Reverse Denoising Processes
  3. 27.3 Score Matching
  4. 27.4 Noise Schedules
  5. 27.5 Latent Diffusion
  6. 27.6 Text-to-Image Systems
  7. 27.7 Video Diffusion Systems
  8. 27.8 Diffusion Transformers
Forward Diffusion ProcessesDiffusion models are generative models built around a simple idea: learn to reverse a gradual corruption process.
10 min
Reverse Denoising ProcessesThe forward diffusion process gradually transforms data into noise.
8 min
Score MatchingDiffusion models can be understood from multiple mathematical viewpoints.
8 min
Noise SchedulesA diffusion model needs a rule for how noise increases during the forward process.
9 min
Latent DiffusionEarly diffusion models operated directly in pixel space. A model generated images by iteratively denoising tensors such as
8 min
Text-to-Image SystemsText-to-image generation aims to synthesize images from natural language descriptions. A model receives a prompt such as:
8 min
Video Diffusion SystemsVideo diffusion extends image diffusion from still images to moving sequences. Instead of generating one image, the model generates a sequence of frames that should remain visually coherent over time.
9 min
Diffusion TransformersEarly diffusion systems used convolutional U-Nets as denoising networks. U-Nets worked well because images contain strong local structure, and convolutions efficiently model nearby spatial relationships.
8 min